Home

Master Thesis semanticSBML a Tool for Creating, Checking

1. G0 1234567 bio is a2 Annotation http www geneontology org G0 1234567 1 0 a3 Annotation BioModels BIOMDO000000001 model is The annotation objects created in the example above can be used to add anno tations to a model ae addAnnotation al The adding of the annotation will modify the libSBML model instance To save the changes persistently the model has to be written to a file on the harddisk by using libSBML functions 3 2 6 Annotation GUI In comparison to the first version of the annotation GUI visible new features are the qualifier modification widgets and annotation resources as hyperlinks Figure 17 shows a screenshot of the new annotation GUI with numbers indica tors for the different features that will be explained in the following legend Legend of Figure 17 1 Choice widget with a list of biological and model qualifiers 2 Hyperlink to a world wide web location that will open a external browser 3 Clicking the change push button will set the qualifier of the annotation modAnnotation that was chosen see 1 44 semanticSBML File Help Main ID gt SBML Config Annotate BIOMD0000000064 Bio is 2 Model BIOMD0000000064 ARES eS E y Models is KEGG Compound C00004 y Teusink2000 Glycolysis ID Teusink2 adh Wy Compartments Ww cytosol ID cytosol earch by Name y ektracellular region ID extracell
2. Rule Size of Circle for Alternative Rule Az3 Y Marvin None 3 Jannis None use alternative Jannis None 3 Marvin None use alternative Figure 9 Resolution of circular rule definitions Base Merge view models Merge view init 0 rung collisionMenud circleMenud init_0 slotResolve_merged elemval2strd slot_remove_circled slotAlwaysResolve left valuesDict rulesDict slotAlwaysResolve right slotResolve left slotResolve right slotResolve keep finish Figure 10 View classes of the merge GUI based on the SBMLmerge algorithm list of duplicate entities can not be accessed through the interface In the ini tialization the function find collision is called If it returns None there are no duplicate entities that have conflicting values If the function find collision returns an object duplicate entities that contain conflicting values were found The returned object contains the duplicate libSBML elements as well as a dic tionary see Appendix A of the values of the entities with flags that show if the values are in conflict The values are then ordered by their biological im portance and converted to string representations elemval2str if necessary The dictionary keys contain descriptions of the values This information is used to create the user interface widgets When the user resolves the conflict by pressi
3. dx doi org 10 1038 nbt926 Luciano J S PAX of mind for pathway re searchers Drug Discov Today 10 937 942 2005 URL http dx doi org 10 1016 S1359 6446 05 03501 4 Hucka M et al The systems biology markup language sbml a medium for representation and exchange of biochemical network models Bioinfor matics 19 524 531 2003 Strmbck L amp Lambrix P Representations of molecular pathways an eval uation of sbml psi mi and biopax Bioinformatics 21 4401 4407 2005 URL http dx doi org 10 1093 bioinformatics bti718 Home site for the systems biology markup language sbml 2007 URL http sbml org retrieved December 2007 Extensible markup language xml URL http www w3 org XML Finney A amp Hucka M Systems biology markup language Level 2 and beyond Biochem Soc Trans 31 1472 1473 2003 URL http dx doi org 10 1042 Hucka M Finney A Hoops S Keating S amp Novre N L Systems bi ology markup language SBML level 2 Structures and facilities for model definitions http sbml org specifications sbml level 2 version 3 release 2 sbml level 2 version 3 rel 2 pdf Novre N L et al Minimum information requested in the annotation of biochemical models miriam Nat Biotechnol 23 1509 1515 2005 URL http dx doi org 10 1038 nbt1156 Teusink B et al Can yeast glycolysis be understood in terms of in vitro kinetics of the constituent enzymes testing biochemistry Eur
4. lined regions dense red areas can be recognized The areas represent clusters of MIRIAM annotations and models The best observable cluster blue out line enlarged image in Figure 27 contains 5 models that describe the glycolysis BioModel 42 37 61 13 64 12 71 38 63 39 The second cluster that can easily be recognized green outline enlarged image in Figure 28 contains 9 models that describe the reaction networks of the mitogen activated protein kinase BioModel 9 40 11 41 14 41 10 42 26 43 28 43 30 43 27 43 31 43 5 of these 9 models originate from the same work 015 Curio1998 purineMetabol 070 Holzhutter2004 cyte Metabolism 051 Chais ee amen 2 Metabolism sinPho lation 071 Or o iam u 064 Teusink2000 ysis 063 Galazzo1990 7 iy kinetics 061 Hynne2001 coly sis 042 Nielsen1998 ees 013 Poolman2004 Hi Morrison1989_Folate ele 056 cd CellCycle 106 Yang2007 ArachidonicAcid 049 Sasarawa2005 MAPK Figure 27 Close up of a cluster of models that describe the glycolysis and similar biochemical networks The dashed line indicates that part of the image was removed 145 Wang2UU7 XII induced Ca Oscillation 031 Markevich ffid MAPK arderedMM kinases 027 Marlizvich2004 MAPE oxdrrc d MM 030 Markevich2005_MAPE_AllRandomElementary 029 Markevich2001 MAPE phosphoRandamElementarr 026 Muthevici2004_MAPE_ vadeved Ele nie nay uw 014 Levchenlo 000 MAPE Seaffeld NV Levchenka 2ffif MAPK noSea
5. On the inserted text from the input form or from the file a regular expression is applied that filters all KEGG reaction identifiers This means that another SBML file can be used as input In the second step all reactions are presented in a human readable form The list can then be modified by using the back push button This will show the first view again with the filtered list of identifiers in the input textfield widget To create a proper model the compartments of the reactions can be specified From a list of Gene Ontology 25 identifiers one entry can be selected with the help of a dropdown box widget The creation may fail In this case an exception is raised by the creation class that will be presented in a error message box If the creation was successfull the view class will send a signal that a new document was created This will trigger a function that adds the new document to the main view The new document can now be treated like any other externally created SBML model Implementation Even though a view in a strict sense only represents existing models a model creation view was created see Figure 7 It contains three important functions makeInitMenu slotNext slotKegg2Sbml In creation of the class instance the makeInitMenu is called This creates the first view Two slots can be called from that view The first opens a file open pop up widget slotInfileBrowse and the second creates the second view slotNext The file in
6. is qualifier is recognized all annotations with other qualifiers are not recognized Please annotate your model or wait for a new version of semanticSBML SBMLcheck depends on annotations gt semantic dependency check 3 1 0 error According to their annotations the reactions v8 and v9 are identical error According to their annotations the species S1 and S2 are identical error According to their annotations the species C1 and C2 are identical warning 10 reactions could not be checked due to missing annotations gt conservation constraint check 0 1 0 warning 21 reactions not checked for conservation relations due to missing annotations overlap check 0 0 0 physical value check 0 0 0 rules check 0 0 0 Hilerzecutingfs a 0 O Model BIOMD0000000090 1 List of Compartments 2 external ID c0 annotation not supported 3 cytosol ID c1 annotation not supported 4 mitochondria ID c2 annotation not supported 5 List of Species 6 S04 ex ID sul ex 7 EtOH_ex ID eth_ex 31 Hm ID Hm 32 List of Reactions 33 vi1 ID vi annotation not supported 34 v13 ID v13 annotation not supported 35 v2 ID v2 bad annotation 52 vLEAK ID vLEAK annotation not supported 53 v12 1D v12 annotation not supported lt lt lt Annotation Menu gt gt gt I list elements without annotations la list elements and their annotations d lt ELEMENT_NUM gt delete annotation a
7. 41 Levchenko A Bruck J amp Sternberg P W Scaffold proteins may bipha sically affect the levels of mitogen activated protein kinase signaling and reduce its threshold properties Proc Natl Acad Sci U S A 97 5818 5823 2000 42 Kholodenko B N Negative feedback and ultrasensitivity can bring about oscillations in the mitogen activated protein kinase cascades Eur J Biochem 267 1583 1588 2000 43 Markevich N L Hoek J B amp Kholodenko B N Signaling switches and bistability arising from multisite phosphorylation in pro tein kinase cascades J Cell Biol 164 353 359 2004 URL http dx doi org 10 1083 jcb 200308060 74
8. A modifiable dropdown box and a non modifiable dropdown box Conflicting values of e g type bool can not be modified in comparison to floatingpoint numbers The user has the choice to resolve the problem by using the values from the resolution widgets or by using the values of either of the elements for this one entity In addition to that the user can chose to solve every conflict by using one of the models values for all conflicting entities There is also a option to keep both entities however it is disabled by default and has to be enabled in the configuration tab Choosing to keep both entities may result in a erroneous model After resolving all conflicts in duplicate entities it is checked if circular rule definitions exist rules that are defined by itself directly or indirectly To resolve this a graphic of the circular rule and its alternatives is drawn and al ternatives can be chosen with push buttons see Figure 9 Implementation In its first development stage the merging of models in se manticSBML uses the SBMLmerge algorithms These algorithms were modified in order to create an interface based merging algorithm that could be used in a model view controller software design pattern The SBMLmerge merge algo 23 semanticSBML File Help Main ID gt SBML Config Merge Merge Description Value A Conflict Value B Name NADH a NADH Id NADH ok NADH Annotation KEGG C00004 nadh KEGG C00004 nadh ChEBI 169
9. ATP cytosol APS cytosol PAPS cytosol SO4 cytosol EtOH cytosol ADP cytosol H2S cytosol CyS cytosol NADH cytosol NAD cytosol AcCoA cytosol OAH cytosol 02 mitochondria S1 mitochondria S2 mitochondria ADP_mit mitochondria ATP_mit mitochondria ATP cytosol_2 APGI rutnenl 211 PAPCIcutnenl 211 fanAlrutnen 211 Figure 30 Simulation of the merged respiratory oscillation model Four oscil lating concentrations of selected species are shown The values are identical to those of the identical species in the second cell see Figure 31 5 Conclusion In the first phase of the development a fully functional release of semanticSBML was created The release includes a clean installation of seman ticSBML The GUI was completed by adding the creation and merging of SBML models functions Section 2 3 A CI was developed which includes a batch pro cessing ability that enables a user to automate the functions of semanticSBML Section 2 4 In addition to that a simplified API was introduced that enables the usage of semanticSBML as a external programming library Section 2 2 The functions of semanticSBML in the first development phase are based on SBMLmerge In Section 2 3 2 the GUI to the SBMLmerge merging algorithm was introduced and the problems of the old merge algorithm were shown In the second development phase this
10. J Biochem 267 5313 5329 2000 Hynne F Dan S amp Srensen P G Full scale model of glycolysis in saccharomyces cerevisiae Biophys Chem 94 121 163 2001 Schulz M Uhlendorf J Klipp E amp Liebermeister W SBMLmerge a system for combining biochemical network models Genome Inform 17 62 71 2006 72 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 Sanner M F Python a programming language for software integration and development J Mol Graph Model 17 57 61 1999 Trolltech Qt Cross platform rich client development framework URL http trolltech com products qt Karp P D et al Expansion of the BioCyc collection of pathway genome databases to 160 genomes Nucleic Acids Res 33 6083 6089 2005 URL http dx doi org 10 1093 nar gki892 Vastrik I et al Reactome a knowledge base of biologic pathways and processes Genome Biol 8 R39 2007 URL http dx doi org 10 1186 gb 2007 8 3 r39 Snoep J L amp Olivier B G JWS online cellular systems modelling and microbiology Microbiology 149 3045 3047 2003 Novre N L et al BioModels database a free centralized database of curated published quantitative kinetic models of biochemical and cellular systems Nucleic Acids Res 34 D689 D691 2006 URL http dx doi org 10 1093 nar gkj092 Wolf J Sohn H Heinrich R amp Kuriyama H Mathematical analysis of
11. MergedEntity is reference instead The merging of the BioEntitys MIRIAM annotations is an aggregation of all non identical annotations The aggregation can lead to contradicting annota tions that is two annotations with the qualifier is reference the same database with different identifiers If this problem occurs a flag set to indicate the con flict The user must solve the problem by deleting non applicable annotations If the problem still exist on a resolution attempt of the user a MergeError is raised For the resolution of conflicts that can be solved by a choice The set func 55 tions of the BioEntity BioQuantity and ModelStatement classes are used The set functions do not only set the values of the semanticSBML object in stance but also of the underlying libSBML object instances In addition to that the semantic correctness of the chosen values is checked A semantic error oc curs if the BioQuantiy type is amount and but an initial concentration is set An invalid semantic will raise a MergeError If the resolution of all conflicts is successful a flag is set that indicates that the problem is solved resolution flag Element Types For the merging of elements of different types the following combinations are allowed parameter with species and parameter with com partment The type of the resulting element always prefers the more complex element type species and compartment This is similar to a strategy that is used in th
12. complete databases but also on a smaller scale for single models Since systems biology is already strongly dependant on computational methods the preferred method of merging complex models is also a computer aided method Two im portant preconditions enable this task a common language for expression of models and a method for recognizing biological objects that describe a model biological entities 1 1 Preconditions Language Formats There are a range of systems biology language formats Each was created to express different aspects of a model The most common ones are CellMI 2 Proteomics Standards Initiative Molecular Interaction XML 3 PSI MI Biological Pathways Exchange Language 4 BioPax and Systems Biology Markup Language 5 SBML Besides many design differences 6 the main feature that differentiates SBML from the other language formats is that it supports mathematical statements to describe quantitative models SBML is a widely used format over 120 software tools support SBML 7 SBML is a XML 8 derived and is developed in levels Each level stands for an addition in the expressiveness of the format 9 The current level is 2 which includes the creation of hierarchical models and usage of spacial characteristics The language is specified in the Systems Biology Markup Language SBML Level 2 Structures and Facilities for Model Definitions document 10 which is one of the main sources of this thesis Appendix B
13. file on the harddrive deserialization 70 B SBML base elements The SBML format is a hierarchic format that has in its first level a row of element types that represent the main concepts of the SBML model These elements are referred to as base elements or libSBML elements in this thesis In libSBML the base elements can be accessed from the model instance with the listOf functions In this thesis the SBML elements are written in a slanted font Since the names of the elements also represent biological concepts the element and the concept can not always be differentiated The following description gives and overview of the most important SBML base elements used in this thesis A full description can be found in the SBML specification 10 species The species element represents a physical entity e g a chemical compound like ATP or a protein or protein complex compartment The compartment represents a bounded space in which species are located e g the nucleus the cellwall or the cytosol reaction The reaction element represents a transformation transport or binding process typically a chemical reaction that can change the quantity of one or more species The reaction contains a mathematical statement the kinetic law parameter Not all symbols used in mathematical statements in SBML must be defined by e g species compartment The parameter defines a symbol that is associated with a values However in different models a par
14. first phase the new version of libSBML was incorporated and the GUI was updated to fit the needs of the newly developed algorithms 3 1 Porting to libSBML 3 The porting to the new libSBML was simple since it was the goal of this phase to rewrite the main algorithms Some changes had to be applied to the file management for writing loaded models to the hard drive Functions developed by my colleagues were simplified while leaving the source code intact so that they could be adapted over time and did not disturb the main functions The parts of the check algorithm that are concerned with the annotation of SBML elements were removed In the future these parts can be replaced by errors that are raised by the new annotation classes The graph visualization algorithm depends on non backwards compatible func tions of the libSBML and had to be disabled 3 2 Annotate One of the main tasks of semanticSBML is the annotation of SBML models with MIRIAM annotations The MIRIAM annotation does not only play a role in the merging of models but could also become a standard for publicly released SBML models 3 2 1 The MIRIAM annotation Introduction The SBML format allows the annotation of elements and of the whole model An annotation can be e g two dimensional coordinates of icons that represent a reaction in a graphical visualization by the popular tool CellDesigner 27 Annotations are optional and their format can be specified by their cre
15. human interaction However in most cases find and replace operations had to be applied by hand All regular expressions that were used were col lected and added as a resource to the project This resource was also used by my colleague for the porting of the web interface Publishing this resource will mostlikely help other developers porting their GUI from PyQt3 to PyQt4 directly applyable regular expressions sed i s QObject QtCore QObject 0 sed i s SLOT QtCore SLOT 0 sed i s QString QtCore QString 0 sed i s QGridLayout QtGui QGridLayout 0 sed i s QDialog QtGui QDialog 0 sed i s QTabWidget QtGui QTabWidget 0 remove extras form previous porting attempts sed i s QtGui QtGui QtGui e sed i s QtCore QtCore QtCore 0 replace import sed i s from qt import from PyQt4 import QtCore QtGui 0 api differences within classes gy c eee does not exist any more qApp setMainWidget gui setMultiLinesEnabled replace manually regular expression may not be exact gt s insertTab S insertTab 32 M1 WARNING unsafe this will mess up the previous changes s insertTab S N addTab M s N message VC X showMessage M1 2001 s setCurrentPage index0f setCurrentWidget M1 s setCurrentPage d setCurrentIndex 1 incomplete this depends on programing style s qApp processEvents Q
16. information was used for the creation of a new merging algorithm The SBML MIRIAM annotation plays and important role in the merging of SBML models and in the creation of models that should be released to the public On this account the MIRIAM annotation manipulation algorithm was updated to fit the current status Section 3 2 The update was achieved by a complete rewrite of the annotation algorithms since the underlying library libSBML added a native support for MIRIAM annotations as well as for the fact that a more flexible design could be achieved The new annotation algo rithms also introduced a new API and an improved GUI for the manipulation 67 Concentrations Volumes and Global Quantity Values 0 ES orcas lass va Punumornununa S0 arriuyiusu_ A APS cytosol_2 PAPS cytosol_2 SO4 cytosol_2 EtOH cytosol_2 ADP cytosol_2 H2S cytosol_2 CYS cytosol_2 NADH cytosol_2 NAD cytosol_2 AcCoA cytosol_2 OAH cytosol_2 02 mitochondria_2 S1 mitochondria_2 S2 mitochondria 21 ADP_mit mitochondria_2 ATP_mit mitochondria_2 Figure 31 Simulation of the merged respiratory oscillation model Four oscil lating concentrations of selected species are shown The values are identical to those of the identical species in the first cell see Figure 30 of MIRIAM annotations In an experiment the new annotation API was used to cluste
17. instances It has functions to add remove and modify annotations Function Description init Read MIRIAM annotations type id metaid and name from libSBML element Only MIRIAM annotatable libSBML elements are allowed as input readAnnotations Internally used function to read all MIRIAM annotations from CVTerms of the inserted libSBML element isAnnotated Return if the element contains MIRIAM annotations addAnnotation Add MIRIAM annotation to the SBML element represented by this class instance In libSBML versions 3 0 2 the adding of identical annotations will create two separate identifiers As a result of this work libSBML prevents this in later version The function also checked if a CVTerm with the same qualifier already exists and add annotation to this CVTerm this functionality was discontinued since libSBML already provides it modAnnotation Modify the qualifier of an annotation remAnnotation Delete an annotation from libSBML element and resynchronized the in ternal list of annotations with libSBML element unsetAnnotations Delete all annotations of this element This function is not present in the current libSBML but might be added in future versions It was created due to the behavior of libSBML lt 3 0 2 described in addAnnotation getAnnotations Return a list of Annotation class instances These instances should be used to add or remove annotations getQuerys Return a list of the name id and me
18. libSBML was updated The MIRIAM annotation manipulation as well as the merging algorithm was rewritten The concept of annotation qualifiers was integrated For the annotation and merging of models independent abstractions of systems biology mod els were developed The merge abstraction is used for a better detection and resolution of conflicts in matching biological objects Experiments were conducted to show the functional efficiency of the new algorithms as well as to show its possible uses Acknowledgment I would like to thank Wolfram Liebermeister for his enthusiasm and for being the best tutor I could imagine My girlfirend Jana for her patience and for giving birth to our child Nila Edda Klipp and the Computational Systems Biology Group especially Jannis Uhlendorf Anselm Helbig and Marvin Schulz for their work on semanticSBML you created this too Ulf Leser for sharing his independent view on our problems The lib SBML community for driving me mad and helping me all at once My family and friends A special thanks goes to Christian Ehrlich and Jonathan Schuld for proofreading my thesis Selbstandigkeitserklarung Ich erklare hiermit dass ich die vor liegende Arbeit selbst ndig und nur unter Verwendung der angegebenen Quellen und Hilfsmittel angefertigt habe Berlin den 07 Februar 2008 Contents 1 Introduction 1 1 Preconditions uum lcu 3 a duo ES 1 2 Previous Work x 4M bo oo a ee ee RES 1 3 Procedure u
19. libsbml so HAHAHA As emant i cSBML f 755 root sys bindir semanticSBML srcdir semanticsbml gui py f 755 root sys bindir semanticSBML console srcdir semanticsbml_console py f 755 root sys bindir semanticSBML id2sbml srcdir semanticsbml id2sbml py 31 f 755 root sys bindir semanticSBML check srcdir semanticsbml_check py f 755 root sys bindir semanticSBML 2dot srcdir semanticsbml_2dot py f 755 root sys bindir semanticSBML exportDB srcdir semanticsbml_exportdatabases py f 755 root sys bindir semanticSBML reduce srcdir semanticsbml_reduce py f 755 root sys bindir semanticSBML stabilize srcdir semanticsbml_stabilize py lib f 644 root sys pylibdir semanticSBML srcdir semanticSBML py f 644 root sys pylibdir semanticSBML srcdir semanticSBML pyc The script starts with defining variables that will be used later on lines 1 10 Required package meta information is listed in the lines 12 20 Setting the cor rect version number allows an update of the program with the removal of old program code In lines 22 25 the package dependencies are set These can then be automatically resolved by a packet management system Since semanticS BML is dependant on libSBML a binary distribution of libSBML is included in the package lines 27 34 The executable scripts of semanticSBML lines 37 44 are renamed and copied to the bin directory during the installation All other scripts are place into th
20. lt ELEMENT_NUM gt add suggested annotation automatically annotate List of or whole Model s ELEMENT NUM QUERY search and add identifier f ELEMENT NUM DB ID add an identifier directly DB KEGG GO q back commands help dir rec prec play hist q exit you can use ctrl4D win ctr Z to exit i exit closing all documents now 29 The example above shows an interactive CI session The CI is called with a list of commands that are then processed line 9 The session starts by setting the switches to open a file and setting the command queue to the commands c 0 and a 0 The p switch executes the command on startup after opening the file The command c 0 executes a semantic check of the open model number 0 line 31 and is followed by the calling of the annotation view with the command a 0 line 56 The view diplays the complete annotations status of the model lines 58 to 109 followed by the available commands that can be used to manipulate the annotations lines 112 to 118 The call of the annotation view created a nested instance Both instances are exited by using the exit command line 121 The string closing all documents now indicates that the modification status of all open models is checked Discussion The console class is a generic class It is my hope that it will be reused by other developers since it is distributed under the same open licence as semanticSBML 2 5 Beta Release Gen
21. of annotations and different repre sentations of the annotation Function Description init Set the resource database and identifier and qualifier qualifier and qual ifier type eq Equality operator If database and identifier are the same return True __str__ Return the annotation resource as specified by the proposed MIRIAM annotation standard URI ID getName Return a human readable string representation of the annotation if the resource can be found in the internal database or an empty string if the resource can not be found getURI Action If possible return a hyperlink to find the referenced element on the world wide web setQualifier Set the qualifier of the annotation libSBML encoded the biological qualifiers in numbers between 0 and 7 and model qualifiers between 0 and 2 and qualifier types between 0 and 2 Both numbers and string rep resentations e g hasPart are allowed as input for the qualifier and the qualifier type If the qualifier is not recognized an error is raised setLink Set the database and the identifier of the Annotation class As input for the database a URI e g http www geneontology org or name e g Gene Ontology are both accepted The input allows setting a flag upon which the insertion of unknown databases known ones are specified in listofresources xml raise an error If the identifier pattern is known for the inserted database and the inserted identifier does not matc
22. of the SBML base elements left to the semanticSBML datastructure right can reference other mathematical statements the SBML function definitions base element Copies of the referenced function definitions are attached to the mathematical statements Similar to the storage of units this can lead to a redundancy The BioEntity BioQuantity and the ModelStatement classes are all derived from a base element that stores information about the type and the model they are derived from This is needed for the correct initialization of all attributes Compare For a pairwise comparison of all MergeEntitys the entities of each MergeModel are traversed The compassion itself is executed by a function of the class BioRelations The BioRelations class holds a triangular score matrix for all biological BioModels qualifiers matched against itself is vs is is vs has part The MIRIAM annotations of each element are compared with the MIRIAM annotations of the other element If two annotations are found to be equal by identity Annotation eq operator see Section 3 2 4 or by belonging to the same group in the internal database the score of the quali fiers is looked up in the score matrix and returned e g is vs is returns 10 All scores are added up and returned as an overall score If the overall score is higher than 0 an attempt is made to create a tuple containing both elements or adding one of the e
23. to reference the same database 5 World wide web location of the database 6 URL that can be combined with an identifier to create a hyperlink referring to a description of the annotation 7 Space separated list of libSBML elements that can be an notated using this database 8 Regular expression pattern that all identifiers of this database must follow The pattern is used for a basic check of the annotation identifiers 9 Closing of the resource tag The listofresources xml file is located in the semanticSBML subdirect which is located in the home directory of the user that installed semanticSBML and thus in a easily accessible location The users of semanticSBML are encouraged to edit the file to fit their needs 3 2 4 Implementation API The implementation of the annotation algorithm follows closely the concept pre viously described in Section 3 2 2 The objects described in the concept match exactly the classes of the annotation algorithm The class diagram can be seen in Figure 16 The interface functions and some internal functions will be shown in the following listing 39 The Annotation class represents a single MIRIAM annotation It consits of methods to get and set the variables database identifier qualifier qualifier type When setting these variables multiple checks are executed to verify their correctness The class uses the external file listofresources xml In addition the class provides functions for comparison
24. view class a submenu is created The source code of the class is shown in the next example class Id2Sbml_view QWidget def __init__ self help lt lt lt ID gt SBML gt gt gt e lt IDi ID2 gt Enter a List of KEGG Reaction Identifiers q exit this menu cc CustomConsole e self slotNext_l insert list q self exit exit help run Also user input can be returned directly without connecting it to a function input CustomConsole raw_input Are you sure you want to do this y n The raw_input function of the CustomConsole class is used instead of the na tive Python raw_input function since its input is captured and can be replayed Just like the GUI the CI interface consists of a main class and views for each of the functions of semanticSBML The views can be found in the console_views module and correspond in their make up to the GUI views Interface Usage semanticsbml console py h usage semanticsbml console py options lt sbml xml files options h help Show this help message and exit q CQ cmdqueue CQ set the command queue v verbose let functions crash p play play stored command queue on start semanticsbml console py q c 0 a 0 p release_25September2007_sbmls curated BIOMDO000000090 xml E loading identifier database 49143 entries loaded in 0 00 03 197326 loading reaction database finished in 0 00 0
25. 08 id not found ChEBI 16908 id not found Compartment cytosol ok cytosol Initialamount 0 39 0 39 0 33 InitialConcentration 0 39 0 3 M 0 33 keep both use A use choice use B always use A always use B Figure 8 Conflict resolution in the old semanticSBML GUI based on SBMLmerge rithm is designed to merge models parwise To enable a merging of an arbitrary number of models the merge algorithms is first called with two models and then subsequently with the resulting merged model and one of the remaining mod els The interface to the merge algorithm was developed in cooperation with my colleague to contain the following functions Function Description _ init__ Initialize the merge class with two models as input find_collision Find duplicate entities that contain conflicting values resolve_collision Resolve the conflict of the duplicate entities values find_circle Find circular rule definitions and also resolve the circular rule definition problem by adding extra parameters deleteunused Delete unused elements in the created model finish Return the newly created model To merge two models the Merger class is initialized with two model instances The user cannot see or influence the compare and initial matching process A 24 semanticSBML Ele Help Main ID gt SBML Config Merge Merge Choose one of the following Rules to remove a circular Rule definition 1
26. 11 species The conflicts were resolved by choosing the preselected concentrations The two remaining match ing elements had next to a conflicting initial concentration also a conflict in their annotations 63 The species High energy phosphates the element names are used in this description of the Teusink model matched the species ATP in the Hynne model The analysis of the annotation conflict revealed that the element of the Teusink model had two resource KEGG compound identifier C00002 and ChEBI identifier ChEBI 15422 with the qualifiers has part that were identical to the resources in the Teusink model with the qualifiers is On checking the Teusink model a species P was found that would biologically match the species High energy phosphates in the Hynne model However the P species was not MIRIAM annotated and was thus not recognized as a match This problem could be resolved by manually matching the two biologically matching species In the current state of semanticSBML the manual matching of elements does not update depending elements and thus excluded this solution As an alter native solution a user could add the correct annotations to the Hynne model species P before the merging is performed This conflict resolution bypasses the species ATP since no similar species could be found in the Teusink model This is a hint that there are more severe problems between the concepts of the two
27. 5 092594 opened BIOMD0000000090 lt lt lt semanticSBML main menu gt gt gt list all loaded models lt FILENAME gt open a model lt DIR gt open all models in the directory without arguments last used lt MODEL_NR gt display check results for a model lt MODEL_NR gt annotate a model with database identifiers Panor 28 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 87 88 89 90 91 e export a svg image of a model m lt MODEL_NR1 gt merge 2 or more models by inserting a list of models s save model s v save model s as r lt MODEL_NR gt remove model about about this software i ID gt SBML Generate SBML files from Database Identifiers commands help dir rec prec play hist q exit you can use ctrl D win ctr Z to exit Hezecuting c 0 gt semantic check 0 0 0 gt annotation check 0 10 1 warning Reaction v2 no Annotation recognized warning Reaction v10 no Annotation recognized warning Reaction v14 no Annotation recognized warning Reaction v4 no Annotation recognized warning Reaction v5 no Annotation recognized warning Reaction v6 no Annotation recognized warning Reaction v7 no Annotation recognized warning Reaction v15 no Annotation recognized warning Reaction v17 no Annotation recognized warning Reaction v18 no Annotation recognized information Model contains elements with missing or unrecognized annotations Currently only the
28. 9 02 22 rdf syntax ns xmlns bqbiol http biomodels net biology qualifiers xmlns bqmodel http biomodels net model qualifiers gt lt rdf Description rdf about metaid_0000075 gt lt bgbiol is gt lt rdf Bag gt lt rdf li rdf resource http www geneontology org G0 0005829 gt lt rdf 1i gt lt rdf Bag gt lt bgbiol is gt lt rdf Description gt lt rdf RDF gt lt annotation gt lt compartment gt The example above shows the SBML element compartment The element is annotated with a reference to the Gene Ontology identifier for cytosol The MIRIAM annotation in human words states The element compartment is the identifier GO 0005829 cytosol of the database http www geneontology org Gene Ontology The table below explains the important sections of the ex ample and introduces the terms that will be used in the following description Line Term Example 2 Entity compartment 6 Qualifier bqbiol is 8 Database Data Type http www geneontology org 8 Identifier GO 0005829 For a complete description of the example the SBML and RDF specifications should be used The originally proposed term data type will be referred to as database The word annotation will be used synonymous for MIRIAM annotation 3 2 2 Concept The main concept of the annotation algorithm is that is a simplified abstraction of a SBML model see Figure 15 The main idea is that a model consists of ele ments t
29. IP suga Aai Qsquawayzajdno juoniab 1usuinuDisse7peniutgpt Qsajdno 1196 ajnazp DIapourqygss1e21 quawajaabuawzp Qsajdnopabsayyyaim syun7 eqojb dal Japour jugs qi lapopjabiayy suoejayo g siapowabaa quawalqabsayy n p3 j0523121 Ju037 u2 Uia je jS a potu Qquawayersjapowmaajosas Augo anjos D anuenpolq j0534 Qs jos21 Qiu2usieis apop s1eda1d Quonea2o1 sgeda4d pAmnuengorgTasedadd Qsuoreiouux s4edaid Q24eda4d pias gpn PIJUOJUM IUIS poU paa osagi juo3 Aunuenpolq ipjuojur Anuenporq panjosay131144037A4143o gq 1nju05ur Aanugorq iusurs 3pab4a i Figure 23 Class diagram of the improved merging algorithm The Merger class is the interface to external classes A model is represented by the MergeModel class It contains a row of MergeElements Each MergeElement consists of one BioEntiy one BioQuantiy and one ModelStatement The MergeTuple is a smart container for MergeElements The BioRelations class provides functions that are used in the context of the merged model and that use external resources 53 SBML semanticSBML Compartments MergeEntity Species Parameters Reactions BioEntiy Unit definitions dp m BioQuantiy Function definitions Initial assignments Rules ModelStatement Compartment types Species types Constraints Events no supported Figure 24 The figure shows the mapping
30. Master Thesis semanticSBML a Tool for Creating Checking Annotating and Merging of SBML Documents Falko Krause krause_fQmolgen mpg de February 7 2008 Free University of Berlin Department of Mathematics and Computer Science Bioinformatics Dr Wolfram Liebermeister Max Planck Institute for Molecular Genetics Computational Systems Biology Group Prof Dr Ulf Leser Humboldt University Berlin Knowledge Management in Bioinformatics Abstract The System Biology Markup Language SBML is a common language for expressing biochemical sets of reactions that are accompanied by mathe matical statements such as kinetic infomation The program semanticSBML provides the systems biology community with the abil ity to integrate merge and annotate models with MIRIAM annotations User interfaces are provided on multiple levels application programming interface API console interface CI graphical user interface GUI This work aims to enable a full support of SBML level 2 version 3 for the merging of models including mathematical statements and the ma nipulation of MIRIAM annotations including annotation qualifiers It is based on the previous work of the Computational Systems Biology Group Max Planck Institute for Molecular Genetics SBMLmerge In its first development phase it extended SBMLmerge with a cross platform GUI and CI for all existing algorithms as well as a simplified API In its sec ond phase the underlying library
31. a mechanism for autonomous metabolic oscillations in continuous culture of saccharomyces cerevisiae FEBS Lett 499 230 234 2001 Riverbank PyQt Python bindings for trolltech s qt application framework URL http www riverbankcomputing co uk pyqt Epydoc Automatic API documentation generation for python URL http epydoc sourceforge net Kanehisa M The KEGG database Novartis Found Symp 247 91 101 discussion 101 3 119 28 244 52 2002 Ashburner M et al Gene ontology tool for the unification of biol ogy the gene ontology consortium Nat Genet 25 25 29 2000 URL http dx doi org 10 1038 75556 Easy Software Products EPM ESP package manager URL http www epmhome org Funahashi A amp Kitano H CellDesigner a process diagram editor for gene regulatory and biochemical networks BioSilico 1 159162 2003 RDF XML syntax specification revised 2004 URL http www w3 org TR rdf syntax grammar Berners Lee T Fielding R amp Masinter L Uni form resource identifier URI Generic syntax URL http www gbiv com protocols uri rfc rfc3986 html Berners Lee T Uniform resource locators URL a syntax for the expression of access information of objects on the network URL http www w3 org Addressing URL url spec txt 73 31 The BioModels qualifiers URL http www biomodels net index php s Qualifiers retrieved December 2007 32 Degtyarenko K et al ChEBI a database and onto
32. ameter can be represented by a species or compartment and vice versa initial assignment An assignment of an initial value of an entity e g species compartment parameter can be archived by either setting the value attribute of the element or by using a mathematical expression the initial assignment An initial assignment always refers to another element rule The libSBML element rule contains a mathematical statement that is used to define dynamic properties of an entities e g species compartment libSBML parameter value There are three different types of rules algebraic rule rate rule and assignment rule The rate rule and assignment rule always refer to another element The algebraic rule rep resents a mathematical statement that has the general from 0 f x and can thus refer to many elements 71 References 1 2 11 12 13 14 Snoep J L Bruggeman F Olivier B G amp Westerhoff H V Towards building the silicon cell a modular approach Biosystems 83 207 216 2006 URL http dx doi org 10 1016 j biosystems 2005 07 006 Nielsen P F amp Halstead M D The evolution of CellML Conf Proc IEEE Eng Med Biol Soc 7 5411 5414 2004 URL http dx doi org 10 1109 IEMBS 2004 1404512 Hermjakob H et al The HUPO PSI s molecular interaction format a community standard for the representation of protein in teraction data Nat Biotechnol 22 177 183 2004 URL http
33. are compared a list of matching elements is created interfaces for modyfieing the match list are provided matching elements can be in conflict interfaces for conflict resolution are provided if there are no conflicts or all conflicts are resolved the internal representation of the model is translated back into a new sbml model semanticSBML ui resources ML annotate Anne ML annotate Anne ML annotate Annc ML annotate Annc ML annotate Annc e semanticsbml gui Trees Indices e semanticSBML Generated by Epydoc 3 Obetal on Fri Dec 28 0401 40 2007 http epydoc sourceforge net Figure 4 Source code documentation with Epydoc 2 2 Application Programming Interface API Documentation The Python programming language provides native methods for source code documentation These native methods can be extended by ex ternal tools to generate a clearly structured source code documentation that includes hyperlinks between classes text highlighting and input output type description among other things The description of input and output of func tions in Python is especially important since Python uses a dynamic type system usually referred to as duck typing types of variables are assigned dynamically by detection of the type of the first inserted value of the variable The dy namic type system requires a proper source code documentation Without the documentation a user of the semanticSBML API would have to guess input and out
34. ase 25th September 2007 release http www ebi ac uk biomodels 4 1 Clustering The MIRIAM annotation in SBML allows an automated identification of bio logical entities A SBML model can be defined by the biological entities it uses to describe a biological phenomenon These two facts were combined to create an automated method of finding similar models The datasource of the experiment was the complete BioModels database 25th PEC Teac ante MO s alae v SAN Glyanlyaie Figure 26 Clustered BioModels database the rows represent models and the columns MIRIAM annotations of model elements Two clusters are outlined An enlarged view of the blue outlined cluster can be found in Figure 27 and a green outlined cluster in Figure 28 September release curated models only The semanticSBML annotation API was used to extract the MIRIAM annotations from each model A Python script was written that generates a matrix of the occurrence of MIRIAM annotations 61 in each model The matrix contains a 1 if an identifier occurs in a model and a 0 if not The qualifiers of the annotations were ignored This matrix was then loaded into MATLAB 36 and clustered with the clustergram function which uses a hierarchical clustering algorithm The clustered matrix was visualized and manually analyzed Multiple groups of models were recognized Figure 26 shows an overview of the complete database with two regions outlined in blue and green In the out
35. ator as long as they follow the external XML standard RDF 28 As mentioned in the introduction Section 1 MIRIAM was created as an effort to ensure the quality of a model and enable a fast entity recognition MIRIAM itself is a proposed framework of rules that consits of two parts The first part describes the syntax and semantics a model description should follow The second part is an annotation scheme This annotation scheme can be applied to SBML elements when encapsulating it into a RDF element The RDF format is used to create semantic statements about an object using a subject predicate object expression The subject is in this case a libSBML element a biological entity The object is an external resource that holds a reference description of the entity The external resource is given by a pair that consits of an URI 29 that is joint with the symbol and an identifier string to from a URL 30 The URI representing a data resource that provides a description of a biological entity which can be found with the identifier string The predicate that describes the relationship between the subject and the object is given by 35 PUNE BioModels qualifier elements 31 The following example shows a MIRIAM annotation in SBML the example is part of a SBML document denotes the rest of the document lt compartment metaid metaid_0000075 id cytosol size 1 gt lt annotation gt lt rdf RDF xmlns rdf http www w3 org 199
36. ction depend on programming style 17 Table of Contents Everything Modules eemantieSRMT hace wawe Y D Everything All Classes PyQt4 QtGui QMain Window PyQui QiGui QMainWindow PyQt4 QtGui QPaintDevice P PyOt4 OtGui QWidget Rende ML Model Model ML Model NeedF i hide private ames no frames Trees Indices Help Module Hierarchy Class Hierarchy Module Hierarchy semanticSBML CustomInteractiveConsole Interactive Console a generic Interactive Console it is used for the console userinterface of semanticSBML semanticSBML Model Interface to all functions in the semanticSBML core for one document actions semanticSBML annotate e semanticSBML base views Console User Interface Views Baseclasses so far only a basecalss for merge e semanticSBML config semanticSBML database semanticSBML docmanager Document View Manager actions on multiple documents e semanticSBML gui views Graphical User Interface Views this module contains all views of a semanticSBML model the views are connected to the controll classes the connection should only be made through the SemanticSbmlGui doc class or kegg2sbml class for one document actions and the Merger class for multi document actions semanticSBML merge 2 Merging of Multiple SBML Models this module enables the merging of sbml models to merge two models an internal representation of the model is created the elements of the internal representatin
37. cument can execute necessary functions for e g updating the view The document man ager class manages multiple documents It creates and connects views to the documents To enable one document and multiple document actions in a user interface the document manager provides functions to show the number of active selected documents and to change the activation state of multiple documents It also takes care of the safe closing of documents to avoid data loss 2 3 Graphical User Interface GUI The GUI is provided by a main class which creates the main window In the main window the user can load documents and a list of loaded documents can be seen By selection of a document item in the main view documents can be activated An activation or deactivation will enable or disable the push buttons that trigger the creation of views Each of the functions of the core code are represented by a view A view is created in a new tab of the main tab widget The tab has the same name as the function key and may contain another tab widget In the following the views for the merging and creation of models algorithms will be described 2 3 1 Model Creation Interface Usage In semanticSBML models can be created by the insertion of a list of KEGG 24 reaction identifiers The GUI view provides this functionality in a two step process see Figure 6 The first step is an insertion of list of KEGG reaction identifiers This can be done either by file or directly
38. d Terms B SBML base elements 69 70 71 List of Figures o DO 0t C l2 EH NN h2 h2 b2 L2 RP RP LEE KP Ao HB CF L BOdocoOo Iccu cr bw 25 26 27 28 29 30 31 Comparison of two glycolysis models Model view controller design pattern in semanticSBML semanticSBML merge concept model abstraction Screenshot Sourcecode documentation with Epydoc Simplified class diagram of the API Screenshot Model creation GUI 2 2 2 222 Model creation class diagram nen Screenshot SBMLmerge merge GUI conflict resolution Screenshot SBMLmerge merge GUI resolution of circular rules Merge GUI class diagram len Console user interface class diagram Screenshot semanticSBML on Ubuntu Linux Screenshot semanticSBML on OS X ee Screenshot semanticSBML on Microsoft Windows New annotation concept a New annotation algorithm class diagram 2 Screenshot New annotation GUI New merge concept cartoon part 1 New merge concept cartoon part 2 o New merge concept cartoon part 3 nenn New merge concept cartoon part 4 o o New merge concept cartoon part 5 o New merge algorithm class diagram Mapping of SBML base elements to semanticSBML merge datas TEUCBULE a IIND ee LE Seo Scr
39. del could be merged directly The original model and the merged model were loaded successfully into COPASI and a time course simulation was conducted for 100 steps The con centrations values of the species were plotted against time Figure 29 shows the result of the simulation for the original model Four non extracellular species were selected that showed oscillating concentration values The Figures 30 and 31 show the results of the simulation of the merged model The same species were selected to show the oscillation in the merged model The concentration values of each species in one of the cell compartments were exactly the same The concentration values in the merged model however differed from those in the original model This is due to the fact that the substance powering the reactions in the cell were used up by two cells in the merged model This means that only half of the substance amount can be used by each cell The experiment proves the functional efficiency of the new merging algorithm to create complete and simulatable models The created model has little sci entific value in itself but it shows an alternative usage of semanticSBML than merging different models into one model Even though the merging algorithm is still incomplete a successful merging could be conducted using only functions of semanticSBML 66 Concentrations Volumes and Global Quantity Values T2 o 00 o Lili lll AB N eo 0
40. delete the MIRIAM annotation In case the aggregated MIRIAM annotations do not represent the merged element 8 Unit string representation 57 vl el cl wT Saa ar ayer Boydsoydsig e 1 seeds E bb 90 0 esuo ET uongemuenuo enu WwawaJeIS 1050343 u0ne2o1 asp 0 L0625 0 ULIAH asp queysu0 gt Bey uopenuexuo enu wwawage s 1050447 uone o1 8 el OT OWL un Ka OT Sour nun voge quasuoy sadAL voge quaauo reddy lt My OT Sot um uonenue uo5 sadAL en el tl ee TT T 2 2 2R Ayquen asoan ason 262007 punoquio 5533 si 26200 punoduo 9 3 si L UT TUNE np FEZZTTIFDIERD Peel TaD TEN 5i Em 52505 punto 5535 9 suonejouu y uone jouuy ied peste sog abia 10N 00 ebuew JON 09 G up lt 4 sanads 19 asom 1epie eiyx3 epo 210 lt e usnap senads 050347 u asconp Sylepow 48490 sob TEZZUTSSH2 TH a A armen mus nn wo syuawajg ui 2513 wouy squeusp3 YIM 9B42j4 suogejouug wp 05 DO Is o2A D OOOTAUISN L sisA ooA 5 TOOT UAH Em esp o bLDE S 0 de queysuo gt Er One 4820 entr losoyis uone o1 uf aemneseayca qr 4epnije2e axe Queue BEE ION Byuo gt was al deu wu 3j qr esoonib 2105034 per2eds o adaw Wasopuewes my Figure 25 The figures shows a screenshot of the the new merge algorithm GUI The sc
41. dels The semanticSBML merge algorithm tries to realize this concept of the merge process however some changes had to be made for an enhanced usability Since this description in this section is a simplification Section 3 3 2 will describe the actual implementation 3 3 2 Implementation The merging algorithm is interface based A class diagram of the merge algo rithm can be seen in Figure 23 The interface to the merge algorithm is provided by the Merger class The Merger class is initialized with a list of libSBML doc ument instances Translation The first step in the merging of the model is the translation of the SBML model into the semanticSBML datastructure Figure 24 shows the basic mapping of SBML base elements to the semanticSBML datastruc ture The SBML base elements compartment species parameter and reac tion are viewed as mergable biological entities and are therefore translated into MergeEntitys Their values and subelements are stored in either the BioEntiy the BioQuantity or ModelStatement class The three classes are stored in the MergeEntity class along with other general properties Identification The BioEntiy class is used to identify the biological object It stores an elements MIRIAM annotation using a Annotations Element see Section 3 2 Description The BioQuantiy class describes an element One if its attributes is type of the BioQuantity The type in the case of a physical biological en tity species determin
42. e global library directory of Python Discussion Distributing a binary version of libSBML was welcomed by its de velopers and a creation of a separate libSBML package was requested A binary distribution package for libSBML currently does not exist for Linux platforms EPM supports most of the mayor Linux distribution and in future development it is aspired to also build packages for other Linux distributions 2 5 3 Cross Platform Ability The installation and the functional efficiency of semanticSBML was tested on the platforms Ubuntu Linux see Figure 12 OS X see Figure 13 and Win dows see Figure 14 While working flawlessly on Linux and OS X the graph visualization with Graphviz did not work on the Windows platform File Edit Help Main ID gt SBML Configure Merge Checked Models Annotate BIOMD0000000061 BIOMD0000000064 4 ATP consumption Model BIOMD0000000061 W List of Compartments Current Annotations o ATP consumption ID vconsum Acetaldehyde flow ID voutACA empty Y Acetaldehyde out ID vdifACA empty amp W Adenylate kinase ID vAK amp W Alcohol dehydrogenase ID vADH amp W Aldolase ID vALD cyanide flow ID vinCN empty Cyanide Acetaldehyde flow ID vlacto empty amp W Ethanol flow ID voutEtOH In amp W Ethanol out ID vdifEtOH 4 e GO GO 0006200 None delete amp W cytosol ID cytosol e y ex
43. e initial creation of MergedEntitys that is if one of the elements merged elements contain more information this information is used in the cre ation of the MergedEntity Manual Matching The manual matching contains two basic cases An el ement can be removed from a tuple or an element can be added to a tuple The removing can result in the destruction of the tuple The adding can also result in the destruction of the original tuple and it can result in the creation of a new tuple If an element is added to a tuple where an element of the same model already exists the element is removed from the tuple The manual matching causes a structural change in the merged model which means that all objects depending on any of the involved elements e g that have a mathematical equa tion pointing to the element have to be updated The first implementation to resolve this problem was a global recreation of all MergedEntitys While it solved the problem it proved to be computationally too expensive Translation The final step is the translation of the semanticSBML model back into a SBML model The translation is only executed if all MergedEntitys are conflict free or have resolved conflicts If this is not the case a MergeError is raised listing all elements that are still in conflict The translation starts with a creation of an empty SBML model It continues with the collection of all SBML base elements that are attributes of MergedEntitys rules initial as
44. e the argument can not be valid The implementation has the disad vantage that many weak annotations can outvote a strong annotation Taking the BioModels database as the primary source of MIRIAM annotated models an outvoting in randomly merged models was not observed With more sources for MIRIAM annotated SBML models the current solution might turn out to be insufficient and has to be revised On this account the matching function was placed into the BioModels class which was created on purpose as an extra class to contain critical algorithms separate from other algorithms The severity of conflicting BioEntity and BioQuantity attributes is much lager than of that of ModelStatement attributes A user must be properly instructed to understand this in a public released version of semanticSBML A severe con flict can mean that a merging of certain entities will result in the creation of a faulty model Since the user can manually decide which entities should be merged it was decided that a warning of severe conflicts is a better solution than to prevent the merging of entities with sever conflicts 60 4 Experiments The following experiments were conducted to prove the functional efficiency of the new algorithms as well as to evaluate the usefulness of the concepts used in semanticSBML The experiments also show alternative uses of semanticSBML All models used in the experiments originate from the current release of the cu rated BioModels datab
45. ed while entities in the extracellular space were merged In a successful merging of the oscillation of the cell should be seen in both cells with identical concentrations for each substance A simulation was performed with COPASI 4 2 23 development and the results were analyzed The model was prepared by creating a copy of the original model and renam ing the copied model Since the manual modification of dependant elements is currently not possible in semanticSBML the models were prepared in such a way that the compare step see Section 3 3 2 would directly yield the desired matches On this account all MIRIAM annotations except for those of extra cellular entities including the compartment extracellular space itself of the copied model were removed The removing of the annotations was done using the semanticSBML annotation algorithm The original model and the copy was 65 Concentrations Volumes and Global Quantity Values 64 49 6 509 20 40 60 80 ATP APS PAPS SO4 EtOH ADP H25 ai E NADH NAD AcCoA OAH F 02 S31 S2 ADP mit ATP mit Figure 29 Simulation of the Respiratory Oscillation Model Wolf et al 2001 in COPASI with four representative species that show oscillating concentrations over time then merged Since the models were identical except for their MIRIAM anno tations there were no conflicts in the matching elements and the mo
46. ee Appendix A It is then deserialized during the instantiation of the singleton class Similar to the command queue a history of commands is kept that can be copied to the play cmdqueue Implementation Safety There are two methods to exit a console inter face session a local exit for a single instances and a global exit for all nested instances In some cases an on exit function is needed for e g the safe clos ingof documents The on exit function can be set for each instance of the main 27 BR POW ANOAOPWNR m w On PWNR N eL o o 100 class by using setOnExit This function is called during each exit attempt and may prevent the exit Implementation Integration In the following example the module is loaded and a dictionary of commands and function as well as a help text is created and inserted into the CustomConsole class The command loop is then started with the run function from semanticSBML CustomInteractiveConsole import CustomConsole self locals 21 self listFiles List Models i2s Id2Sbml_view ID gt SBML d self openDirectory Open Directory self _help lt lt lt semanticSBML main menu gt gt gt 1 list all loaded models d lt DIR gt open all models in the directory without arguments last used i2s ID SBML Generate SBML files from Database Identifiers cc CustomConsole self locals self help runO By invoking the creation of the Id2Sbml
47. eenhot New merge algorithm GUI Overview of clustered BioModels database Cluster of glycolisis models o Cluster of mitogen activated protein kinase models Simulation of the respiratory oscillation model Simulation of the merged respiratory oscillation model cell 1 Simulation of the merged respiratory oscillation model cell 2 1 Introduction The field of systems biology tries to explain complex relationships in biological systems Its focus is in the integration of information 1 to discover emergent properties that could not be revealed with other methods A common approach in systems biology is to create a model of a metabolic process that has been researched by many scientific groups with the help of physical experiments As the field of systems biology grows so does the amount of information it gener ates in particular the amount of models it has created Models are expressed and published in various ways from databases with custom datastructures to proposed exchange formats to verbal descriptions accompanied by mathemati cal statements The vision of systems biology is to not only study the behavior of selected bio chemical networks but the biochemical network of a whole cell or even a whole organism To reach this goal an integration of the generated models is needed This integration can be achieved not only on a large scale e g the integration of
48. em biology scientists that create and simulate models in a mostly tool driven process The CI provides an interface that can be easily automated without knowledge of the Python programming language The console interface is geared to the interactive Python console It was devel oped since no similar Python module was available It is designed to aid users with little programming knowledge in Python or in general creating simple automated tasks with semanticSBML Implementation Concept The console interface functionality is provided by the CustomInteractiveConsole module see Figure 11 It consist of two classes the main class CustomConsole and a singleton datastorage class Singleton datastore The main class can be instantiated with a dictionary containing commands as keys strings and function pointers as values When 26 CustomConsole command_dict on_exit_fkt exit help local Singleton_datastore verbose exit recording cmdqueue play queue init 0 rung play raw input setOnExit setCommandQueued historyd Figure 11 The console interface is based on the CustomInteractiveConsole module The module consists of two classes the CustomConsole and the Singleton_datastore The Singletone_datastore provides a singleton class that stores values that can be used in nested instances or that should be applied on all existing instances of CustomConsole the run function is called a loop is started that prompt
49. ent rule rate rule algebraic rule event and the model itself The support of libSBML elements is dependant on the listofresources xml file and can be extended by the user This will be described in the following para graphs Predicate It is able to modify all current BioModels qualifiers Model Qualifiers is The modelling object represented by the model component is the subject of the referenced resource For instance this qualifier might be used to link the encoded model to a database of models isDescribedBy The modelling object represented by the component of the encoded model is described by the referenced resource This relation might be used to link a model or a kinetic law to the literature that describes this model or this kinetic law 37 unknown The qualifier is unknown This is not a part of the BioModels qualifiers but of the libSBML It is needed since a qualifier is mandatory for a annotation Biological Qualifiers is The biological entity represented by the model component is the subject of the referenced resource This relation might be used to link a reaction to its exact counterpart in KEGG or Reactome for instance hasPart The biological entity represented by the model component includes the subject of the referenced resource either physically or logically This rela tion might be used to link a complex to the description of its components isPartOf The biological entity represented by the model com
50. enud slotNextd slotKegg25bml0 slotinfileBrowseb Figure 7 Create Model view class functions by the kegg2sbm1 function the creation failed and the error message is displayed in a pop up window 2 3 2 Merge Interface Useage If a user wants to merge models in the semanticSBML GUI he has to select the desired models in the main view and press the merge push button It is only enabled if more than two models are selected Upon the pressing the merge button the modification state of the models is checked If a model is unsaved the merging is aborted and a error pop up appears warning the user to save the model first documents states see Section 2 2 Tf the models are not modified there are two possible results The first result can be that the merging was successfull without user interaction The merged model is then returned right away The new model will appear in the main window having a modified state The second result is that a new tab appears that again contains a tab widget in which the merging takes place see Figure 8 The merging is executed on models pairwise If the models that are to be merged contain du plicate entities that have conflicting values the attributes of the each entity will be displayed in a vertical list In between the two entities the merged attribute values are displayed If the values are in conflict a conflict resolution widget is displayed see Figure 8 There are two possible conflict resultion widgets
51. eral There are two methods to install semanticSBML The first method is the source installation and the second is a packaged installation The installa tion method depends on the operating system The creation of the distribution packages will be described in the Sections 2 5 1 and 2 5 2 For the beta release the INSTALL file was updated to describe the installation process in detail The READEME was updated to contain a basic description of the current functions of the program Clean Up To prepare the official release the root folder of the project had to be cleaned up from test scripts and library scrips in development Since se manticSBML is a project that includes source code that has been developed by my colleagues and myself moving and removing of files had to be done carefully and with the agreement of the corresponding developer The inclusion criteria of executable scripts was that they had to at lease return a list of switches Publication To complete the release the Sourceforge project site was updated and screenshots of the program were uploaded The project was renamed to semanticSBML and the distribution packages were uploaded After the update of the institute homepage of the project an announcement was made to the libSBML mailing list 2 5 1 Source Installation To use semanticSBML on the Windows Microsoft Inc or OS X Apple Inc platform it has to be installed with the Python installation script from the package Python D
52. erent element types and the manual choice of elements that should be merged It includes methods for a simple recognition of severe conflicts and semantic checks of the user selected conflict resolution The representation of the merge process is more transparent due to the merging of an arbitrarily number of models at once and an improved visualization of element values In the current state the program delivers a framework that needs further im provement Missing features for example are the resolution of circular rule definitions which was included in SBMLmerge and circular compartment in clusions Further semantic checks are needed which can only be discovered with a deeper understanding of the SBML format The program does not yet support all SBML element types An improved implementation of the resolution of the dependency problem in which only the depending elements are recreated was started but could not be finished with this work SemanticSBML currently compares even complex elements like mathematical statements and units by string comparison While this might seem ineffective experience showed that it yields some success especially if standard units are used and models are derived from each other An appropriate solution for the units comparison would be a conversion of the units or a standardization of units before the comparison The libSBML developers announced that functions for a unit conversion are currently integrated in the
53. erge multiple models duplicate entities are identified and a list of du plicate entities is generated merge tuples The list of merge tuples can be manipulated by a user since the duplicate detection may not be correct or may not reflect the wishes of the user From the merge tuple a new entity can be cre ated However the merge tuple may contain conflicts since e g different models may make different statements about the duplicate entities e g the entities have different initial concentrations The user must resolve all conflicts e g choose the correct initial concentration After the resolution of all conflicts a merged model can be created The resolution of conflicts and the creation of the final model is associated with many difficulties An example for such a difficulty would be the fact that a SBML model must contain unique identifiers for each element A detailed description of the merge concept can be found in Section 3 3 1 4 Experiments Databases There are a number of databases that support the export of SBML models BioCyc 17 Reactome 18 JWS online 19 These databases however do not support MIRIAM annotations in their current version The largest source of curated models which includes MIRIAM annotations is the BioModels 20 database It contains models from diverse sources that are added to the database after a curation step in which among other things the MIRIAM 14 annotations are added This makes it the b
54. erity of a conflict In addition to that the abstraction allows a simple reuse of functions that are needed during the merge process 46 SBML soe secan nn HEEL Figure 18 Merge Concept In the first step large grey number of the semanticSBML merging algorithm the SBML document is translated into semanticSBMLs own abstraction 3 3 1 Concept Next to the model abstraction the the mergin process in semanticSBML uses further concepts that will be introduced in this section In the following class names written in typewriter font are used along with artificial names for datastructures LibSBML elements are written in slanted font The first step is the translation of a SBML model to a semanticSBML model see Figure 18 step 1 the steps are marked with large grey numbers On this account the libSBML base elements see Appendix B compartment species pa rameter reaction these elements are considered as mergable biological entities are translated into MergeEntitys The MergeEntity class can be viewed as meta SBML element since it contains all of the attributes the different types of SBML elements can contain In some cases an attribute can also be a SBML base element The MergeModel stores all MergeEntitys of one libSBML model The translation is repeated for all models that should be merged see Figure 19 step 2 All elements in each model are compared pairwise with the elements of matching type in the other models see Fig
55. es if it is an amount or a concentration SBML property hasOnlySubstanceUnit in the case of a compartment it is the dimension of the compartment e g the nucleus is a three dimensional compartment whereas the cellwall is a two dimensional compartment The class also stores reaction participants reactand product and modifier for reaction elements In addition to that the class stores the unit and location of an element The location at tribute contains a pointer to another MergeEntity of the type compartment The unit is stored as a custom dictionary type The storage of units is redun dant in comparison to SBML units are a base elements and can be predefined but easier to use since standard units are resolved and can be used along with custom units Statements The ModelStatement class holds the largest amount of informa tion A part of the attributes can be summarized as simple attributes e g the initial amount float of a species or the flag for reversible reactions bool The simple attributes are either of type float or bool The class also stores mathematical statements In the case of a reaction the mathematical state ment is the kinetic law and in the case of compartment species and param eter it can be a rule or initial assignment The mathematical statements are stored as copies of libSBML element instances The mathematical statements 52 Quoniuaquonoun4u ieuie Quuawubissyye Quuawubissyyeniupab Qywwawayeysyep
56. escribed in Section 3 The GUI and CI are currently used as a template for the creation of a web interface which is currently under development by a colleague 2 1 Porting to Qt4 Preliminary Work The creation of the GUI was started during my research internship The development described in this thesis picked up the work where the research internship left off It was decided during the research internship that Qt should be used as a widget see Appendix A toolkit Qt provides a stable cross platform library for the creation of GUIs Qt is written for the C programming language For Python an interface is provided by the PyQt 22 developed by Riverbank library Porting Mandatory At the beginning of the masterthesis the support of Qt version 3 the current version at the time of my research internship was dis continued due to the release of Qt version 4 The new version contained non backwards compatible changes in the application programming interface e g functions were renamed and thus required the porting of the existing source code The final decision to port to Qt4 was made when the official distribution webpage of Qt no longer offer binary installation files for Qt3 Procedure Since this project is based on the PyQt and not directly on the Qt library the porting tools that were provided by Trolltech could not be used Instead a series of regular expressions were written that could partly be applied 16 PWN a without
57. est source for real life examples of models that use the SBML format The BioModels database was used not only in the experiment as a source for models but also during the whole development of semanticSBML Experiments To exemplify the functional efficiency and the potential uses of the new algorithms three experiments were conducted In the first experi ment a clustering of the BioModels database was performed to exemplify a fully automated method for finding similar models with the aide of the MIRIAM an notation API In the second experiment the merging of the previously introduced glycolysis models using semanticSBMLs new merge algorithm was attempted The third experiment consisted of the merging of a single celled model of an autonomous metabolic oscillations in continuous culture of Saccharomyces cerevisiae by Wolf et al 21 into a two celled model 1 5 Organization of this Document Special terms that a reader of this thesis should be aware of can be found in the Appendix A An overview of the most improtant SBML elements can be found in the Appendix B SBML elements play an important role in Sections 3 2 3 3 and 4 Function and variable names are written in typewriter font This thesis describes each of the development steps in detail Whereas Section 2 has its main focus on the implementation of a fully functional re lease of semanticSBML Section 3 has its main focus on the concepts of the new annotation and merge algorithm wi
58. ffald 000 068 Curienz003 Me Tlu_ssniluesis BAA an io TN 1 Rrartnm REAZT 13h 023 Markesich2004 MAPK khosphafrandomE emertary Figure 28 Close up of a cluster of models that describe reaction networks around the mitogen activated protein kinase The dashed line indicates that part of the image was removed 62 The clustering is a demonstration of an alternative usage of the semanticSBML annotation API The experiment was conducted with rather vague assumptions To achieve better results only annotations with the qualifier is should be used Furthermore it is recommended to choose a more suiting clustering algorithm However even with these loose assumptions the clustering shows the potential of an easy access to MIRIAM annotations The clusters that were presented in this experiment could easily be found by a manual comparison but if the amount of available models rises a manual search for similar models could become a very time consuming task The clustering of models deliver a great alternative to the manual search for similar models 4 2 Analysis of Merging Two Glycolysis Models The example in the introduction Section 1 shows two models that described similar aspects of the glycolysis of Saccharomyces cerevisiae In this experiment it is attempted to merge the glycolysis model of Hynne et al 13 BioModel 61 and the glycolysis model of Teusink et al 12 BioModel 64 with seman ticSBML The merging process will be a
59. h this pattern an error is raised see _checkIdPattern checkIdPattern This function checks if a regular expression pattern for identifiers of a given database can be found listofresources xml If a pattern is found the inserted identifier is matched against the pattern If the pattern matches the function returns True if it does not match the function returns False 40 Annotaion database identifier qualifier qualifierType lisofresources xml init 0 eq 0 str oO setQualifierd setLinkd getNamed getURlActiond checkldPatternd 0 1 AnnotationCreationError AnnotationDatabaseError manage Annotations Element annotations type init_0 addAnnotationd 3 remAnnotation modAnnotationd getAnnotationsd unsetAnnotationsd addAnnotationAutomaticd getQuerys isAnnotated getSuggestions readAnnotationElements hi manage Annotations_Elements_Model annotationsElements fi init_0 getElements getNumNotAnnotatedElements remAnnotationElement _readAnnotationElementsd Figure 16 Class diagram of the improved annotation algorithm The Annotation represents a single annotation The Annotations Element repre sents a SBML element The Annotations Elements Model represents a SBML model 41 The Annotations_Element class represents one annotatable object in a model a biological entity libSBML base element It is a container for Annotation class
60. hat can be annotated The type of an element is defined by an attribute and not by the element itself like in SBML In the new annotation concept a model consits of a collection of elements and elements consit of a collection of annotations Thus the model depends on its elements and elements depend on its annotations However an annotation can be used independently from and element and an element can be used independently from a model The construc tion of the model and element depend on the SBML whereas the annotation can be created independently from SBML The annotation was constructed this way on purpose to enable a possible reuse outside of semanticSBML 36 Annotations_Elements_Model Annotations Element Annotation Figure 15 The concept of the new annotation algorithm Each level of the abstraction used for the annotation algorithm can be used independently The Model and the Element level depend on SBML whereas the Annotation level is SBML independent 3 2 3 Features The basic idea of the RDF format was introduced in Section 3 2 1 The main features of the algorithm are introduced by listing the available values that the subject predicate and object of the RDF annotation object can have Subject The new annotation algorithm supports annotations to the following libSBML elements see Appendix B species compartment reaction parame ter assignm
61. ion the resolution of circular rules is repeated until all problems are resolved Some choices may lead to loops in the resolution Discussion A compromise had to be made between the functional efficiency and the implementational effort The conflict resolution algorithm has a num ber of problems that limits its usability Some of the limits are its pairwise approach in merging the inability to handle entities that seem identical but that are marked as non identical its inability to recognize biological facts like the location of a physical entity and the fact that most of the merging process is hidden for the user These limits will be addressed in the new merging al gorithm that is presented in Section 3 3 The circular rule definition algorithm and its user interface is very hard to understand by a user not familiar with con crete algorithms Since it is one of the features of SBMLmerge it was integrated into the user interface For an actual use it needs further improvements The implementation of the user interface for this algorithms helped to understand the functions of SBMLmerge in detail and to analyze its shortcomings 2 4 Console Interface CI The console user interface provides a user interface that is not dependant on a X Window System however it is still dependant on Qt due to signals send from the document manager One of the main features of the CI is its batch processing ability The desired audience of semanticSBML are syst
62. ist file for semanticSBML definition of variables prefix usr exec_prefix usr bindir exec_prefix bin datadir usr share docdir datadir doc semanticSBML 1ibdir usr lib pylibdir libdir python2 4 site packages mandir usr share man srcdir home foreach MPG semanticSBML trunk sbmlmerge main package information product semanticSBML copyright 2007 Computational Systems Biology Group Max Planck Institute for Molecular Genetics vendor Computational Systems Biology Group description Create Check Annotate and Merge SBML Documents description this package includes libsbml 2 3 4 using xerces with python bindings version 0 9 3 readme srcdir README license srcdir COPYING format deb requires python2 4 Yrequires python2 4 qt4 requires python soappy requires graphviz requires libxerces27 HHHHHHHHLIbDSOML 1 755 root sys libdir libsbml so libsbml 2 3 4 so f 644 root sys libdir libsbml a usr local lib libsbml a f 644 root sys libdir libsbml 2 3 4 so usr local lib libsbml 2 3 4 so f 644 root sys pylibdir libsbml libsbml py usr local lib python2 4 site packages libsbml libsbml py f 644 root sys pylibdir libsbml libsbml pyc usr local lib python2 4 site packages libsbml libsbml pyc f 644 root sys pylibdir libsbml pth usr local lib python2 4 site packages libsbml pth f 644 root sys pylibdir libsbml so usr local lib python2 4 site packages
63. istuils Installation Difficulties The installation requires that all dependencies on external libraries libSBML Qt PyQt Graphviz SOAPpy are satisfied before the installation is started All versions semanticSBML up to 0 9 3 are depen dant on libSBML 2 4 this is again dependant on Python 2 4 libSBML 2 4 30 SCOONODOBRWNHERE 5 Ww N will not work with newer versions of Python Installing Python and libSBML is straightforward since both exist as binary distributions for Windows and OS X The dependencies on PyQt4 and Qt4 however is problematic on Windows Qt4 can be installed with a binary installer that will also install an open source GNU compiler Since Riverbank does not provide a binary installer for PyQt4 depending on Python 2 4 the user has to build it by hand This can be done using the compiler provided by Qt4 Unfortunately a config file of Qt4 on which PyQT4 is dependant is missing a variable at the time of writing that has to be added by the user manually The detailed instructions can be found in the INSTALL file 2 5 2 Debian Package To create a debian binary installation package the software packaging program EPM 26 created by Easy Software Solution was chosen It was also chose for its ease of use It enables a uniform method for building binary software packaged for UNIX Linux systems The building of a software package with EPM requires the creation of a list file The following shows the l
64. lements to an already existing tuple If both elements are in a tuple one of the elements is added to the best matchin tuple The attempt of adding an element to a tuple can fail if the tuple contains an element of the 54 same model with a higher match score In addition to that the match score of the elements location compartment element is looked up and compared with the location score of a competing element The compartment score is always the divisive score if competing elements exist since it is important for the structural identity of an elements in the merged model Matching elements are stored in a MergeTuple All elements that matched but were not merged retain links to eachother These links are used in following merging process to obtain a list of similar elements The discussion in Section 3 3 4 will address the matching problem further Differences to the Concept The concept of the merging process in Sec tion 3 3 1 described that after the compare step Figures 19 and 20 steps 3 5 the manual manipulation of the MergeTuple list would follow Figure 20 steps 6 Only after these steps the conflict resolution Figure 21 steps 7 and the generation of MergedEntitys Figure 21 steps 8 follows However in the im plementation there is no separation between the updating of the MergeTuple list and the merging of tuples The implementation continues with the gener ation of an randomly merged model The random merged model provides the
65. libSBML and will be an inherent 59 part in future version of the library In a discussion with the developers of CO PASI 35 it was discovered that the comparison of mathematical statements is also needed in COPASI It was agreed that the source code for the comparison of mathematical statements would be provided by the COPASI developers The identification of duplicate objects consits of two steps The first step is the recognition of identity and the second is the generation of a similarity score The similarity score is generated by a rather naive algorithm It uses a score matrix to enable a comparison of annotations with different qualifiers e g an notation of modell ATP is version of URI1 ID1 compared to annotation of model2 ATP is URI1 ID1 The score matrix was intentionally not in cluded in this thesis since no combination of qualifiers other than is for both annotations was found to be more meaningfull than other combinations and has a higher value in the score matrix It can be argued that only elements with identical annotations and the qualifier is should be recognized as equal This argument however would not allow elements that only have weak annotations e g is version of to match any other element The assumption made in the implementation is that if the argument would be would be true every biological entity can be referenced with an is qualifier However since no database is complet
66. logy for chemical entities of biological interest Nucleic Acids Res 36 D344 D350 2008 URL http dx doi org 10 1093 nar gkm791 33 Huffenberger M A amp Wigington R L Chemical abstracts service ap proach to management of large data bases J Chem Inf Comput Sci 15 43 47 1975 34 3DMET A database of three dimensional structures of natural metabolites URL http www 3dmet dna affrc go jp 35 Hoops S et al COPASI a COmplex PAthway SIm ulator Bioinformatics 22 3067 3074 2006 URL http dx doi org 10 1093 bioinformatics bt1485 36 The MathWorks MATLAB r2007a URL http www mathworks com products matlab 37 Nielsen K Srensen P G Hynne F amp Busse H G Sustained oscillations in glycolysis an experimental and theoretical study of chaotic and complex periodic behavior and of quenching of simple oscillations Biophys Chem 72 49 62 1998 38 Helfert S Estvez A M Bakker B Michels P amp Clayton C Roles of triosephosphate isomerase and aerobic metabolism in trypanosoma brucei Biochem J 357 117 125 2001 39 Galazzo J L amp Bailey J E Fermentation pathway kinetics and metabolic flux control in suspended and immobilized saccharomyces cerevisiae En zyme and Microbial Technology 12 162 172 1990 40 Huang C Y amp Ferrell J E Ultrasensitivity in the mitogen activated protein kinase cascade Proc Natl Acad Sci U S A 93 10078 10083 1996
67. m of element annotations In Section 3 2 the MIRIAM annotation in SBML will be introduced in detail Figure 1 shows an example of two models of the glycolysis one from Tuesink et al 12 and the other from Hynne et al 13 Even though both models describe almost the same aspects of the glycolysis the recognition of duplicate objects can not be done by e g string comparison of entity names The article describing MIRIAM states We believe their the MIRIAM rules application will enable users to i have confidence that curated models are an accurate reflection of their associated reference descriptions ii search col lections of curated models with precision iii quickly identify the biological phenomena that a given curated model or model constituent represents and iv facilitate model reuse and composition into large subcellular models This thesis shows methods and examples that realize the expressed believes of the authors 14 mark additions made by the author of this thesis 11 View Model GUI Cl view classes base SBMLmerge classes S x N Controller Qt classes Figure 2 Model view controller pattern in semanticSBML 1 2 Previous Work In its previous work the Computational Systems Biology Group Max Planck Institute for Molecular Genetics developed the basis of semanticSBML a pro gram called SBMLmerge 14 SBMLmerge is able to annotate SBML m
68. ment ModelA H1O ModelA Ethanol ModelB F6P ModelB F16bP Merged Model ModelA ModelB Merged ATP ATP ATP Merged Glucose Glucose gluc H20 Ethanol F6P F16bP The main widget of the merge GUI is a toolbox widget The toolbox widget consists of a list of vertically arranged tabs that extend when the tab header is clicked on Inside a tab widgets that each display an element are arranged horizontally The Figure 25 shows a screenshot of the merge GUI The screen shot is composed of four separate screenshots which were combined for a better overview Legend of the Figure 25 1 The header is showing which model is located in which column 2 Symbols representing the conflict status of the element 3 Header of the element displaying the element name The header is color coded for the different types of libSBML elements In this case the blue highlighting represents a SBML species 4 Drop down box widget that contains a list of similar elements with which the element could also be merged Just below it a similar drop down box widget is located with a list of all elements that this element could be merged with The lists only contain elements that do not belong to the same model 5 Push button to remove the element from its current tuple In this case pushing the button destroys the tuple and deletes the merged element 6 Aggregated MIRIAM annotation with hyperlink 7 Push button to
69. models The second pair of matching elements with conflicting annotations were the ele ments Triose phosphate Teusink model and Glyceraldehyde 3 phosphate Hynne model Similar to the problem in the paragraph above the Teusink model had two annotations with the qualifier has part matching the annota tions with the qualifier is found in the Hynne model In addition to that the species Triose phosphate showed that there was a second matching element in the Hynne model Dihydroxyacetone phosphate the generation of a list of similar elements was described in Section 3 3 2 The hyperlinked external resources reveal that the two species matching the Triose phosphate are very similar and a further investigation in the linked resources show that there is a reaction that converts the chemical compounds into one another In the analy sis of the reactions of the models it shows that the conversion reaction is part of the Hynne model but absent in the Teusink model In the current version of semanticSBML this problem is very difficult to solve since an aggregation splitting of elements is not possible The species that were not matched were analyzed It was found that at least two species were not recognized as matching due a problem in their annotations The species Glucose 6 Phosphate and Fructose 6 Phosphate Teusink model should have matched with the species Glucose 6 Phosphate and Fructo
70. n functions directly function A subroutine also known as method procedure or subprogram init new A special method used to initialize classes similar to a constructor in other programming languages biological entity A biological object e g the chemical compound ATP the compartment cytosol the reaction that converts Glucose into Glucose 6 phosphate qualifier A qualifier defines the relationship e g has part is version of be tween two objects e g the relationship between a biological entity ATP and a database entry Reactome entry for ATP A detailed description of qualifiers can be found in Section 3 2 1 widget A graphical element e g a push putton to patch Correcting of a flawed algorithm to wrap Creating a function that has the same input and output of another func tion to raise Raising an exception is also known as throwing an exception signal slot Qt concept of executing functions on user generated events e g the execution of the function that creates a pop up window on the pressing of a push button to port Updating of interface functions of an integrated library For example the function setCaption in Qt3 PyPt3 was renamed to setWindowTitle in Qt4 PyQt3 serialization deserialization In the context of this thesis it is the process of writing an object class instance to the harddrive serialization and then creating the object class instance again from the
71. nalyzed and discussed The merging starts with the comparison of elements and a creation of a ran domly merged model see Section 3 3 2 The following table shows the result of the comparison by listing the number of mergable elements by type for each model and the number of elements that were recognized as duplicates Element Type Teusink Model Hynne Model Matching Elements compartment 2 2 2 species 21 25 13 reaction 18 26 10 parameter 21 0 0 The first thing that can be recognized is that all matching elements have con flicting values Both compartments in each model were matched but contain conflicts in their annotations The annotation resource URI and Identifier in these annotation are recognized as identical However in the Hynne model the MIRIAM anno tations have the qualifier is version of while in the Teusink model annotates the elements with the qualifier is A matching of the elements was possible since the merging algorithm allows the matching of elements with weak annota tions see Section 3 3 2 and 3 3 4 The option is given to chosen which qualifier each of the annotations should have in the merged model The GUI marked the qualifier is as preselected and since this qualifier was chooses it was only needed to press the resolve push button Out of the 13 matching species 11 of the species had conflicting initial con centrations No other conflicts occurred in these
72. ng one of the push buttons a resolve function is called The function resolve collision has multiple conflict resolution strategies e g use values of the left element use the values chosen by the user Each resolution strategy demands a specific input Since the user interfaces all have a similar design the base class Base_Merge_view was created see Figure 10 to unite the input gen eration for the resolve collision function The view class resolve function can return string representations One of the tasks of the base class is the recon 25 version of these string representation The call of the function find_collision is repeated with the steps just described until no more duplicate entities with conflicting values are found Experience showed that circular rule definitions cannot be found while merging two man made models nevertheless the SBMLmerge algorithms can detect and resolve this problem The function find_circle returns a list of rule definition identifiers and their mathematical statement that have a circular definition Al ternatives to each of the rules are given Furthermore a graphical representation of the rule dependencies is returned that should aid the user choosing the cor rect rule The rule identifiers and their mathematical statements are displayed in a table with push buttons that show the rule identifier Next to it the graph representations of the dependencies is displayed Similar to the conflict resolu t
73. ntities is respected ATP in cytosol is not the same as ATP in the mitochondrion merging of multiple mod els at once in SBMLmerge only pairwise merging is possible overview and the controll over which entities should be merged and that mathematical statements are connected to the entities they describe The new merging algorithm works on the basis of semanticSBMLs own abstraction for systems biology models the core abstraction is shown in Figure 3 Merge Concept The semanticSBML abstraction is based on the following con cept a systems biology model consists of multiple biological entitys Biological entities can be identified by MIRIAM annotations Each entity has a biological quantity that describes the entity e g unit location quantity type A model makes statements about biological entities e g the mathematical statement 13 semanticSBML Merge Model Mergable Entity Biological Entity Biological Quantity Model Statement MIRIAM annotation location kinetic law unit quantity value reversible Figure 3 In semanticSBML a systems biology model consists of many mergable entities The mergable entity consist of a biological entity a biological quantity and a model statement Examples of the contents of each of these objects are shown of a kinetic law the value of the amount of an physical biological entity the reversibility of a reaction To m
74. odels with MIRIAM annotations which follow the first draft of the annotations format in SBML Its merging algorithms are able to perform a successfull merging of two SBML models into one model The resulting model has merged mathemat ical statements that were prooven to be functional for simulations In addition to that it contains subrutines to perform a semantic check on a model and create a graphical representation of the model The program is written in the Python programming language 15 The algorithms of SBMLmerge follow the SBML format closely Its design includes a console base user interface directly which is connected to the core algorithms SBMLmerge has several shortcom ings some due to changes in the SBML format and others due to its design Since SBMLmerge is the first tool of its kind it can be viewed as experimental software and many of its shortcomings are addressed in this work SBMLmerge is currently the only publicly available tool that can manipulate MIRIAM an notations using a user interface and performed a merging of SBML models in a automated fashion SBMLmerge is a generic tool that can be used offline an internet connection is only needed during the installation This philosophy is continued in semanticSBML This thesis describes the development of seman ticSBML that is based on SBMLmerge 1 3 Procedure Goals There were two goals for this masterthesis on the one hand to create a fully functional version of the seman
75. of databases by the user It can be argued that this will endorse the use of none accepted databases however semanticSBML addresses a professional audience and will therefore only guide but not restrict a user The design of the internal database has several disadvantages e g a complicated access to the comparison of identifiers and missing relations between identifiers of one database An improvement of the internal database would also improve the annotation process The semanticSBML annotation interface should help the systems biology com munity to accept MIRIAM annotations One of many new vistas that is opened up by an API for the manipulation of MIRIAM annotations is shown in an experiment in Section 4 1 3 3 Merge As already mentioned in the introduction Section 1 semanticSBML uses its own abstraction of a systems biology model The abstraction is based on the idea see Figure 3 that a model consists of biological entities A bi ological entity BioEntity can be identified It is described by a biologi cal quantity BioQuantiy and the model makes statements about the entity ModelStatement This abstraction differs from the abstraction of the SBML format It was created for the following reasons The experiences with SBMLmerge showed that object dependencies in the merging process were not clear The merging of elements with different types should be allowed The conflict resolu tion of SBMLmerge did not differentiate the sev
76. pdating process of the CI it is aspired to create a method for the documentation of the merging process This means that all operations that were conducted during the merging process should be protocolled so that the merging process can be repeated As it was mentioned in the discussion of the new annotation algorithm Sec tion 3 2 7 the internal database needs to be restructured In the current state the functions of the internal database to retrieve information of the identity of different identifiers needs improvement As it was shown in the experiment in Section 4 2 semantic information of identifier relations can improve the merging process String representations of annotations are a good aide in the identifica tion of entities An extension of the amount of data while retaining a fast data access would help the identification of entities by a user and thus improve the the usage of semanticSBML The semanticSBML project is now in its third year of development and has improved the program greatly It is my hope that the development can continue in the future to create a software that will aide the systems biology community to better understand biological life 69 A Frequently used Terms The following list explains special terms that are used in this thesis dictionary Hashtable implementation in the Python programming language module A module represents a file in Python A module usually contains classes but it can also contai
77. ponent is a physical or logical part of the subject of the referenced resource This relation might be used to link a component to the description of the complex it belongs to isVersionOf The biological entity represented by the model component is a version or an instance of the subject of the referenced resource hasVersion The subject of the referenced resource is a version or an instance of the biological entity represented by the model component isHomologTo The biological entity represented by the model component is homolog to the subject of the referenced resource i e they share a common ancestor isDescribedBy The biological entity represented by the model component is described by the referenced resource This relation should be used for instance to link a species or a parameter to the literature that describes the concentration of the species or the value of the parameter unknown The qualifier is unknown See unknown model qualifier Object The database support is not as easy to list as the support for the libSBML elements and the qualifiers It depends on two components of seman ticSBML The first component is the internal database and the second one is the XML file listofresources xml which is included in the program package The internal database provides human representations of identifiers as well as information about the identity of identifiers from different external databases It is used in the search fo
78. provides a short overview of the most important concepts of SBML SBML is becoming a non official standard for the exchange of mathematical simulatable models An important feature of SBML is the availability of a programming library for its access and modification libSBML Object Identification The subject of entity recognition was addressed by the development of the Minimum information requested in the annotation of biochemical models 11 MIRIAM rules MIRIAM was developed as a team effort of systems biology scientists throughout the world and is hoped to 10 GIcX O Glc out Trehalose a Extracellular d er NAD NAD PAT NECS HAP Gic Giycx mM TRIO BPG Sg ADP 2PGA PEP NAD NADH ATP ADP ADP NADH y Ace PYR PEP eis me g j ACAx EtOHx Succinate pn a NADH CHx 0 CQ Figure 1 The figure shows two models of the glycolysis left Hynne 2001 right Teusink 2000 that have common biological objects Some objects can be easily recognized as duplicates blue and green circles others can not red circles MIRIAM annotations can resolve this situation by delivering a method for object identification standardize model curation The goal of MIRIAM was to create a set of rules that ensure the quality of systems biology models A part of the MIRIAM framework describes rules for the creation of a machine readable globally unique descriptions of biological entities These rules were applied to the SBML format in the for
79. put folder of the last file opening is stored over multiple session with the help of the Config class In the second view functions of the API see Section 2 2 are used to display humand readable representations of the reactions us 21 semanticSBML Eile Help Main ID gt SBML Config Input Open a File Containing KEGG Reaction Identifiers Browse Enter a List of KEGG Reaction Identifiers R01061 R00756 X semanticSBML Eile Help Main ID gt SBML Config lt Back Create SBML Reaction Compartment RO1061 2R 2 Hydroxy 3 phosphono oxy propanal E Bur ET G0 0005623 cell v lt gt 3 Phospho D glyceroyl phosphate R00756 ATP D Fructose 6 phosphate lt gt ADP g ee G0 0005623 cell bisphosphate lt Back Create SBML Figure 6 The model creation view enables the creation of models in two steps First step input of KEGG reaction identifiers Second Step visual representa tion of reactions and choice of compartments Gene Ontology identifier ing the id2str function and displays a list of Gene Ontology identifiers using the compartmentIds and id2str_compartment functions By using the Cre ate SBML push button the slot slotKegg2Sbm1 is called It will invoke the kegg2sbml function of the kegg2sbml module If a KEGG2sbm1Error is returned 22 Id2S5bml view idlist init 0 initM
80. put types For this purpose Epydoc 23 Automatic API Documentation Generation for Python was chosen see Figure 4 All functions of semanticS BML that belong to the library interface are documented in detail including input and output parameters Implementation An over view of the API class structure can be seen in Figure 5 The development of the user interfaces required an interface class model class The model class was partly developed by my colleague during the restructuring of SBMLmerge and was refined by myself in the creation of semanticSBML At the same time I developed a module see Appendix A that provided the abstraction of an SBML models for the user interfaces The gen eral interface to SBMLmerge is provided by the class Model The following list shows all important functions of the model API Function Name Description _new__ Initialize the model from either a string of the filename another Model class instance a libSBML document None create a new model 18 save_as Save model under the specified filename using the save function save Save the model as SBML file on the harddisk get libsbml document Return a libSBML document instance get libsbml model Return a libSBML model instance get model as svg Return a graphical representation of the model in the Scalable Vector Graphics SVG format as string This function requires the program dot Graphviz check Run a semantic check on the model and re
81. r annotations The internal database supports the 38 OANODAOPWNR databases Gene Ontology KEGG Reactome ChEBI 32 CAS 33 and 3dmet 34 The internal database was created with SBMLmerge and will thus not be discussed in detail only the functions that are used by the merge algorithm may be mentioned The listofresources xml file provides a list of database URIs names idpatterns etc It is used to show human readable representations of databases create hy perlinks to the databases and do a basic check of the correctness of identifiers The first version of the listofresources xml file was provided by the BioMod els team and was incorporated into semanticSBML since it provides a flexible lightweight method to incorporate new databases for the annotation The file consits of a list of resource elements Each resource represents one database The structure will be explained using the following example resource name EC code uri http www ebi ac uk IntEnz alternateUris http www ec code org location http www ebi ac uk IntEnz action http www ebi ac uk intenz query cmd SearchEC amp amp ec elements assignmentRule rateRule algebraicRule reaction event idPattern d d d d d d d d d d gt Line Description 1 Opening of the resource tag 2 Human readable name of the database 3 Primary URI part of the MIRIAM annotation 4 Alternative URI that can be used
82. r similar models in the BioModels database Section 4 1 Just like the annotation algorithm the merging algorithm was rewritten Sec tion 3 3 A datastructure for the merging was created that has in its center semanticSBMLs own abstraction of a systems biology model For a number of problems strategies were developed to enable a safe and userfriendly merging of models While the merging algorithm is not in a state that it can be released to the public it does deliver a strong framework that can be used to create function merged models Its problem resolution strategies were analyzed in an experi ment Section 4 2 and in another experiment Section 4 3 its functionality was proven 68 6 Further Work The development of semanticSBML is not complete The next important step is the development of a computational lightweight method to update elements after a manual manipulation of the model structure The merging algorithm has to be tested intensely and remaining problems have to be eliminated The testing will mostlikely reveal problems in the semantics of the created models that have to be prevented by integrating further semantic checks The GUI of the merge algorithm has a couple of small problems e g missing scrollbars that have to be corrected in order to provide a pleasant merging experience with semanticSBML The CI delivers a method of automating the merging process The Cl is currently not working with the new merge algorithm In the u
83. reenshot is composed of four screenshots that were combined to present a better overview The dashed lines indicate the borders of the single screenshots The legend to this Figure can be found in Section 3 3 3 58 9 Header of the model statement section The red highlighting indicates that there are conflicting values in this section When the resolution of the conflict was successful it will be highlighted in green 10 Editable drop down box widget that contains choices for the conflicting values The current value is indicated by an arrow icon 11 Resolve push button which will execute the resolution functions Upon pressing the button a pop up window can appear that displays a severe conflict that prevents the resolution of the conflicts 12 Tab header of the next element tuple Clicking the header will open this tab and close the current one 13 Disabled push button that is planned to execute a resolution of all con flicts by choosing only values of one model This feature is still under development 14 Merge push button if all conflicts are resolved pushing this button will create the SBML model and destroy the view If there are still elements with unresolved conflicts a pop up will appear that shows a list of these elements 3 3 4 Discussion The merging in semanticSBML allowes the user more freedoms than SBMLmerge while trying to prevent errors that result from these freedoms New features are the merging of diff
84. rt container that only allows the storage of one entity from each model an aggregation of entities is not allowed It also decides which element belongs in the container A list of MergeTuples is generated see Figure 20 step 5 The user can manually modify this list since a correct matching of the duplicate entities can not be guaranteed see Figure 20 step 6 Even though it is not desired this also enables the manual merging of non annotated models The entities in the MergeTuple may contain conflicting values Most conflicting values can be resolved by choosing a value from a list e g choosing the cor rect initial amount of the ATP see Figure 21 step 7 However conflicts in the BioEntiy determining the identity of the entity or the BioQuantiy describing the entity show that there is a severe disagreement between the entities The resolution of severe conflicts may need more than a simple choice and should often lead to the destruction of the tuple The user can however solve every conflict and thus mark the MergeTuple as resolved From a resolved or non con flicting MergeTuple a combined element is created called the MergedEntity see Figure 21 step 8 The MergedEntity has the same properties as a MergeEntity and has only implementational importance From the list of all MergedEntitys and MergeEntitys a merged SBML model 51 can be created see Figure 20 step 8 The merged SBML model integrates all information of the inserted mo
85. s and the user interface classes were extended to include the merging and model creation functions In addition to the GUI a CI was created Official Release On the completion of the model classes and the views the cross platform ability was verified and the program was intensively tested The project was then officially renamed to semanticSBML The first step in the de velopment of semanticSBML was concluded by an official release New Algorithms The first major change in the development branch was the portation see Appendix A to the current version of libSBML In the current version the libSBML includes native support for the modification of MIRIAM annotations including annotation qualifiers Qualifiers describe the relationship between two objects e g phosphate is part of ATP Since the old MIRIAM annotation manipulation algorithm did not support qualifiers and a new method of editing a MIRIAM annotations was available the algorithm was rewritten For a merging of SBML models MIRIAM annotations are essential as previously mentioned before Changing the annotation interface also meant changing the merging algorithm Furthermore the merging algorithm of SBMLmerge was strongly geared to the SBML format and disregarded the biological context of the model This lead to a complete redesign of the merge algorithm Some features of the new algorithm are a more biologically meaningfull merg ing of models e g location of physical biological e
86. s found in the internal database The annotations that are dispayed in this area can be found using two different methods The first method is an initial search that is conducted when the element is opened getSuggestions The second method is a manually search see 6 45 12 The treeview displays all annotatable model elements The green icon in dicates that the element is MIRIAM annotated the red icon indicate that element is not MIRIAM annotated The tree item displays the value of the name and id attribute of the Annotation element libSBML element If the name is not set but a MIRIAM annotation is available like in the example the internal database is used to diplay a human readable string representation of the first annotation of the element in square brackets getName Expanded nodes display all annotations of an element For adding annotations automatically the nodes of the treeview left hand side of the annotation widget see 12 on the first level e g Species has to be selected This will create a menu on the right hand side of the widget which contains a push putton that starts the automatic annotation of all elements that are child nodes of the currently selected tree node e g all Species elements 3 2 7 Discussion The new annotation algorithm delivers a new API for the manipulation of MIRIAM annotations Its design is flexible and includes all current BioModels qualifiers The algorithm allows the integration
87. s the user for an input and executes the according function if the input matches a keyword from the command dictionary If the user did not enter a keyword all available commands are displayed Since sub menus are needed e g main menu annotate menu it is possible to open a new CustomConsole instance within a running loop The function raw_input provides the ability that only single values can be retrieved This function is also used internally to retrieve user input Implementation Batch Processing To enable a batch processing abil ity a singleton class was created Since multiple independent instance of the main class can be created sub menus single value user input the singleton class provides the ability to share a global queue of commands The singleton class contains two important variables cmdqueue and play_cmdqueue The cmdqueue contains a list of strings keywords If the batch processing is acti vated the list of commands cmdqueue is copied to the play cmdqueue Each time the raw input function is called the function checks if there is an active command queue play cmdqueue If this is the case a value is taken from the play_cmdqueue and returned until all commands on the queue were executed The batch processing mode can be activated on the instantiation of the main class or within a running loop Commands can be added at instantiation or recorded during an interactive session On setting the command queue it is serialized s
88. se 6 Phosphate Hynne model however the match was missed because the Teusink model annotated the elements with a D glucose 6 phosphate ChEBI identifier ChEBI 17665 and KEGG Compound identifier C00668 and 8 D Fructose 6 phosphate KEGG Compound identifier C05345 respectively while in Hynne model the annotations referenced D glucose 6 phosphate ChEBI identifier ChEBI 15954 and KEGG Compound identifier C00092 and D Fructose 6 phosphate KEGG Compound identifier C00085 This problem could be solved by extending the internal database to provide extended infor mation of object relations in this case one object is the parent of the other object 64 All matching reactions had conflicting kinetic laws The choice of the kinetic law needed a deeper understanding of the models and therefore the the prese lected values were chosen Six of the ten matching reactions had conflicting reaction participants reac tands products In four cases the participants were conflicting in the inclusion and exclusion of ATP ADP P as reaction participant In one case the already mentioned problem of the splitting of the Triose phosphate species in the Teusink model into two species in the Hynne model can be seen again The reaction Aldolase same name for both models has in the Teusink model one product and in the Hynne model two products The reaction participant conflict of the matching Glucose 6 Phosphate isomerase Teusink model Phospho gl
89. signments and attached function definitions These elements are directly added to the model After an update of the identifiers of referenced elements of the non merged MergeEntitys the SBML base elements of non merged MergeEntitys and MergedEntitys are added to the model The units of all elements are col lected and their redundancy is removed by a comparison of each of the units They are then added to the model In a final step the function returns the newly created SBML model 3 8 8 Merge GUI The GUI of the new merge algorithm is basically a table in which the rows rep resent elements and the columns represent models The first column of the table represents the merged model All other remaining columns represent the models that should be merged Each row stands for one element of the merged model which is about to be created If there are matching elements their merged element is shown in the first column If an element of a model does not match 56 any other element the first column is left blank In the creation of the merged model the non matching element will be copied with some modifications into the merged model The following table shows an example in which two models ModelA and Mod elB are merged Each model has four elements Two elements from each model are matching an element of the other model ModelA ATP with ModelB ATP and ModelA Glucose with ModelB gluc and two elements in each model are not matching any other ele
90. t of its functions can be found with different names in one of the classes above 43 N ovosoaun m 17 18 19 20 3 2 5 Implementation Integration The following example shows the usage of the model class It is initialized with a libSBML Model instance AnnotaionErrors have to be caught in case the Model has elements with faulty annotations import libsbml from semanticSBML annotate import document libsbml readSBMLFromString open mymodel xml r read try aem Annotations_Elements_Model document getModel except AnnotationError e print e else print aem getNumNotAnnotatedElements The Annotations_Element class is initialized with a libSBML element in this case a species element Like in the initialization of the model the element might raise an exception if the annotations of the element are faulty On a successfull initialization the number of non annotated models is printed to the screen try ae Annotations_Element list document getModel getListOfSpecies 0 except AnnotationError e print e else print ae isAnnotated Annotation instance can be created in different ways In line 17 and 18 iden tical annotations are created using different input In line 17 human readable representations are used in line 18 the URI and the libSBML numbers of the BioModels qualifiers are used see Section 3 2 4 Line 19 creates an annotation for a model al Annotation Gene Ontology
91. tGui qApp processEvents s setMultiSelection 1 setSelectionMode QtGui QAbstractItemView MultiSelection s selectionChanged itemSelectionChanged s QWidget C QWidget 1 s QGridLayout 7 QGridLayout 1 s QTabWidget 7 QTabWidget 1 s QLineEdit QLineEdit 1 gt s QWidget close QWidget close 1 emit does not take tuples anymore gt s emit C 7 s C emit 1 2 s setCaption setWindowTitle 1 s QScrollArea QScrollArea 1 s addChild setWidget change treeview items selection state s S setSelected 7 1 setSelected 2 s insertItem insertItem 0 1 s Qt AlignNCL 2 V QtCore Qt ALignM gt s QFrame C 3 32 QFrame M1 The list is shortened since it is only used to show examples stands for excluded text The regular expressions in the listing above are ordered from directly applyable regular expressions to those that have to be checked carefully by hand The directly applyable regular expressions are written so that they can be copied into a file and used as a shell script that calls sed stream editor for filtering and transforming text All other regular expressions can be copied into vim Vi IMproved a programmers text editor and applied if a correct match is found The regular expressions of this se
92. taid value of the libSBML element These can be used to query annotations from the internal database getSuggestions Get a list of Annotations by querying the internal database If no query is specified as input the function getQuerys is used A switch disables the fuzzy database search then only exact matches are returned 42 addAnnotationAutomatic Check if the element is already annotated using isAnnotated If is not annotated check if an annotation can be found using getSuggestions exact matches only Add annotation s that were found to this element using addAnnotation The Annotations_Elements_Model represents a complete model It is a con tainer for Annotations_Element instances Its functions are limited since it is mainly used as a data container Function Description init Read all elements from a SBML model that can be annotated and create a list of Annotations Elements using readAnnotationElements getNumNotAnnotatedElements Return the number of elements that have no MIRIAM annotations getAnnotationElements Return the list of Annotations Element that can contain MIRIAM an notations available in this model remAnnotationElement Remove an element readAnnotationElements Go through all MIRIAM annotatable elements in a libSBML model and create Annotations Elements In addition to these classes the Merger class of SBMLmerge was replicated for backwards compatibility It is a simple wrapper class Mos
93. th less focus on the implementational details In Section 4 the experiments will be presented The conclusion in Sec tion 5 is followed by and overview of planed enhancements of semanticSBML in Section 6 15 2 Phase I In the first phase the goal was to create a fully functional release of semanticSBML The release includes the library API as well as a graphical and console user interface Since parts of the GUI already existed the GUI had to be extended to include the missing merging annotation and model creation functions see Appendix A For a public release clean up work and updating of user manual files had do be done In addition to that the program was adapted to use the latest version of the Qt library The first official release of semanticSBML contains the following functions Short Name Description display Create a graph visualization of a SBML model id to SBML Create a model from a list of database identifiers check Execute a semantic check on a model e g check for missing MIRIAM annotations annotate Search add and remove MIRIAM annotations for SBML model elements merge Integration of multiple SBML models into one model After the release the source code was branched into two versions a stable release version and a development version The release version was kept for patches of major program flaws one patched release was issued while the main development continued in the branch The new development will be d
94. ticSBML program based on SBMLmerge that would include a graphical user interface GUI and a console user interface CI as well as a revised application programming interface API And on the other the annotation and merge algorithm should be updated to fit the cur rent status of the SBML and MIRIAM formats as well as eliminate flaws of the SBMLmerge algorithms 12 User Interfaces To achieve these goals the existing code had to be adapted to provide interfaces that could be used as a direct programming interface and as an interface for the different user interfaces In my research internship that pre ceded this masterthesis it was my task to design the basis for the graphical user interface The SBMLmerge source code was currently being restructured This lead to an interface based SBMLmerge core that could be used in a model view controller program design see Figure 2 The GUI toolkit Qt 16 developed by Trolltech was chosen It was the goal to create a program that could perform on multiple platforms Until the end of my research internship the interfaces for the check annotate and graph image generator subrutines were created These interface subrutines were collected and wrapped see Appendix A into a single class that could be used as API The class also stores the SBML model and is thus referred to as the model class The model class was used as basis for the development of the user interface views In this thesis the model clas
95. tracellular region ID extracellular Search by Name List of Reactions T metaid_0000051 id s0 Add Database Identifier keca Opened Model merge model 01 Figure 12 semanticSBML running on Ubuntu Linux with the annotation view open 32 Main Config Creare ML Reaction Compartment R00001 Polyphosphate H20 co00563 e 11 lt gt Oligophosphate GO 0005623 cell R00002 16 ATP 16 H20 lt gt 8 e 8 H 16 Orthophosphate CO 0005623 cell A 16 ADP GO 0005623 cell GO 0000015 phosphopyruvate hydratase complex GO 0000108 repairosome GO 0000109 nucleotide excision repair complex GO 0000110 nucleotide e repair factor 1 comple GO 0000111 nucleotide e repair factor 2 comple GO 0000112 nucleotide e repair factor 3 comple GO 0000113 nucleotide e repair factor 4 comple _ GO 0000118 histone deacetylase complex GO 0000119 mediator complex Figure 13 semanticSBML running on OS X Apple Inc with the model cre ation view open 33 Ma semanticSBML DER Add Remove Models mpm Figure 14 semanticSBML running on Windows Microsoft Inc with the main view open 34 3 Phase II The goal of the second phase was to update the core algorithms for modification of MIRIAM annotations and the merging of models Since a new version of libSBML was released during the development of the
96. turn a check results object The semantic check is a function of SBMLmerge and is documented in with SBMLmerge addAnnotationLink Add an annotation link Database Identifier to a specified element if the annotations is incorrect an InvalidAnnotationError is raised addAnnotationLinkAutomatic Add MIRIAM annotations automatically if no annotations were found an InvalidAnnotationError is raised delAnnotationLink Delete an annotation link see addAnnotationLink getNumNotAnnotatedElements Return the number of elements that are not MIRIAM annotated getAnnotationElements Return a list of annotation elements of this model getAnnotationSuggestions Return a list of suggested annotations for an inserted libSBML element getAnnotationQuerys Return a list of query strings that can be used to search MIRIAM anno tations in the internal database for an inserted libSBML element getAnnotation Return the MIRIAM annotations as a list of database identifier tuples of the inserted libSBML element issetAnnotation Return True if the element is MIRIAM annotated This function should not be used but rather getAnnotationStatus 19 getAnnotationStatus Get the status of the MIRIAM annotation of the inserted libSBML ele ment id2str Return a human readable string representation of an inserted database id tuple id2str_compartment Return a human readable string representation for a libSBML compart ment element id2str_reaction Re
97. turn a human readable string representation for a libSBML reaction element compartmentIds Return a list of all libSBML compartment ids databases Return the list of databases used in the internal database for an inserted libSBML element If no input return all databases It is important to note that the model class only includes functions that are working on a single model The merge functions work on multiple models and thus are found in a different class Document Manager SemanticSbmlGui docmgr Merge Model Views Model p Annotate Check Document SemanticSbmlGui doc Figure 5 semanticSBML API The Model class delivers a general programming interface which is extended by the Document Manager and the Document class for the GUI and CI views Specialization of the User Interfaces The user interface module is located in the docmanager module The modules extends the Model class with file management functionalities It contains two classes SemanticSbmlGui docmgr document manager and SemanticSbmlGui_doc document The document class represents a single model and is the direct extension of the model class In additional to the model class functions it provides state variables The state variables are set in the wrapped model functions If a state is modified a signal 20 see Appendix A is send so that all views depending on this do
98. ucoisomerase Hynne model elements hints that the comparison step missed biologically matching elements in this case glucose 6 phosphate and fructose 6 phosphate as discussed above The further analysis of non matching reactions revealed similar problems as the ones already mentioned It should be noted that there were no conflicts between units The analysis of the merging shows that a manual matching process is inevitable Furthermore it can be seen that an aggregation of elements is an interesting field and that semanticSBML needs further improvement The matching of weak an notations was proven to be usefull but it could also be seen that weak matches deserve special attention in the merging process The creation of the randomly merged model revealed missed matches and thus also proved to be usefull The merging of the two models in this experiment might not be possible in the current state of semanticSBML but it showed the strength and weaknesses Ex periments like this will help in the improvement of semanticSBML and possibly also other tools currently in development 4 3 Merging of Respiratory Oscillation Model To verify the functional efficiency of the new merge algorithm see Section 3 3 a model created by Wolf et al 21 BioModel 90 was used The model con stits of an oscillating reaction network within a cell which is powered by two extracellular substances The model was merged with itself in a way that the cell was duplicat
99. ular Species NADH y Glycerol ID GLY W Ethanol ID ETOH Add Extracellular Glucose64 ID GLCo 4 W Succinate ID SUCC Database CQ2 ID CO2 3DMET is EhEBI CHEBI 16526 co2 is KEGG Compound C00011 co2 W Trehalose ID Trh Glicogen ID Glyc no MIRIAM annot PubChem 3306 adh Ska IA r n n am Figure 17 GUI to the new annotation algorithm The legend can be found in Section 3 2 6 10 11 Remove this annotation from the model element remAnnotation Human readable representation of the annotation getName Start a fuzzy search on the internal database for the query inserted see 7 getSuggetsions The results will appear in 11 Editable choice widget with suggestion of queries for a database search getQuerys Drop down box widget with a list of known databases listofresources xm1 that can be used to annotate the current element Clicking this push button an attempted will be made to add the manually constructed annotation to the current element addAnnotation If the inserted identifier does not match the regular expression pattern of iden tifiers of the database _checkIdPattern a pop up appears displaying an error message Add the annotation that is displayed next to the push button to the cur rent annotations of the element addAnnotation The new annotation will have the biological qualifier unknown Annotation that wa
100. ure 19 step 3 The algorithm checks for biological identity with the help of MIRIAM annotations as well as 4T Bicentty _ BioQuantity l Beene feaa Been ean Figure 19 Merge Concept In the second step the translation Figure 18 step 1 is repeated for all documents that should be merged In the third step all mergable elements of each model are compared in a pairwise manner 48 Figure 20 Merge Concept If in the third step Figure 19 duplicate entities are found a MergeTuple is created During the comparison process a list of tuples is build up step 5 that can be modified by a user step 6 49 a ecu seen scr user interaction sen ecu seem scr Figure 21 Merge Concept Tuples of biological entities can contain conflicting values The conflicts must be resolved by user interaction step 7 From a tuple a new entity is created that is the merged entity of all entities in the tuple step 8 50 INCID TS vem 1 z z gt Figure 22 Merge Concept The collected MergeEntitys and MergedEntitys integrate all information of the models that were merged The merging process ends with a retranslation of the semanticSBML model into a SBML model structural identity location of physical biological entities e g ATP in cytosol ATP in mitochondrion If duplicate entities are found they are stored in a MergeTuple see Figure 20 step 4 The MergeTuple is a sma
101. user with valuable information for the manipulation of the MergeTuple list On this account the following paragraph describes the merging process before the updating of the MergeTuple list Merging Before elements can be merged the identifiers of all elements have to be collected and stored in a global list The list is stored in the BioRelations class This is followed by the creation of MergedEntitys from MergeTuples The MergedEntity class is derived from the MergeEntity and contains a copy of all the values of a random MergeEntity of the MergeTuple The MergeEntity class extends the MergeEntity class by boolean variables that indicate con flicting values between elements of the MergeTuple conflict flags dictionaries containing the possible choices for conflicting values as well as functions to gen erate the conflict dictionaries and functions to resolve conflicts Conflict Resolution To generate a conflict resolution dictionary the attribute values of each element are checked for equality For complicated data types the equality is determined by the equality of their string representations The conflict resolution dictionaries contain only values that can be used directly in a merged model This means all referenced element identifiers are updated before they are added to a dictionary Mathematical statements undergo an identifier update with the help of a function of the BioRelations class If the referenced elements are about to be merged the
102. uum 22 RIS auk ee ee Se petu d L4 Experiments 23 222 2 aa n e RD aa data UR 1 5 Organization of this Document o Phase I 2 1 Porting to QUA ein oye duisi Re qp ES 2 2 Application Programming Interface APT 2 3 Graphical User Interface GUI o o o 2 3 1 ModelCreation 22e 2 3 2 Meise cd uou e atatum x xao tue era 2 4 Console Interface CI 2 2 0 2 022 0004 2 5 Beta Release nis otoa ara N bate Joh eh 2 5 1 Source Installation 2 2 2222 22mm nennen 2 5 3 Debian Package ie maaa a 2 0002 eee 2 5 8 Cross Platform Ability 2 2 2222 2 o oo Phase II 3 1 Porting to bSBMI 3 5 ata doo Be de de 3 2 Annotate uc ay bow A A te Eh 3 2 1 The MIRIAM annotation 39 2 2 Concept x A cvy A LS S 3 2 9 ECU Sia Sw XR E FEE G 3 24 Implementation API 3 2 5 Implementation Integration 3 2 6 Annotation GUI e e 3 2 4 DISCUSSION 342 ar ESI EUR NS ues A a e ETE d 3 3 Merge ou dox NATURE MEME Ne RR ee de de ots dl Concepto Ha Sue ok ER esed intr ede te t 3 3 2 Implementation leen 3 9 9 Merge GUT ad AA ee Vg 3 3 4 DISCUSSION Y iua uev nr RE a ee Experiments 4 1 Clustering i e te Geb wage yahi Eon me ede hsb u 4 2 Analysis of Merging Two Glycolysis Models 4 3 Merging of Respiratory Oscillation Model Conclusion Further Work A Frequently use
103. yyas Quuawareisyepiab palunpo Jas pauunjo 136 pazisias pazisia6 Qiuejsuojjas Quueisuo jab Oalqisaaaayias dalg s anay ab Qiseyias Qisegi26 pabieyoras pabieyo196 p mnuengias QA UENO IB 1uauuuBIs eventu quamayersy yeu suun oA azis queysuo3 algisuanad yes aey Amuenb uawa WISP japow7abuaw Adal vuawa a7 11qsqu1 adh wawajgaseg abian Qsauugoigias Q5211u301gi25 Qad januenpias QadAAnuenpisb Quumies Qurumiab puone307135 puonezo7126 PI ___ pasn age sasse 4uoneiouux aui 1241 pasn si SSB asequiep G 24 EUR aun E 5141 TE xuiey 240251357 fenbguonejouuysi Quone07a1e duo Qsquawagaedwos Qaldno jysagasooys QywpyayeysuBsy Opia pas aseqeyeg piwaugpiziapour cts Baos Tuone o red q S212 vorajbu s Qs1u2u2 3195 iiusus 3ua4 Qiusus 3ppe suonejayo g yatwaa pafa w 3103574red squawala ajdno Labaw 1511P1 Quuawajquoneounyiab u inae Quone1ouuyua xueui 34025 ine ad4 Auen L Duis OSUONBIOUNFWERNNNSE abra 4013230 uoneiouwe Tal suone ayoig nuenpoig pafqoog I 0 T T i I abeukw Opps gajdnojzquawalgppe suumebT ajdno 0 juawalzppe opnab Qsuuna Qajdno Lo I3PP sayepipue ajdno 136 Qs2yepipue ajdno ppe Qajdno pias Qajdno 1136 sayepipues ajdnoy aldnoy adaa pimay p Qad Agsiuatus 3125 Qs1ustua 3125 Qajdno jwosjuawayuas OS

Master Thesis semanticSBML a Tool for Creating, Checking

Contents

Download Pdf Manuals

Related Search

Related Contents