CGH-Explorer
Contents
1. 7 8 and so on A segment is called positive if c 1 for genes in the segment and negative if 1 for genes in the segment Step 2 Feature extraction We want to be able to distinguish between segments that are likely to arise from segmentation of normal DNA and those that are not In order to do this we consider two properties of each segment their length L and height H Specifically for each segment we compute the pair L H where L gt 0 is the number of genes in the segment and H gt 0 is the absolute value of the average of the responses of the genes in the segment Step 3 Obtain the null distribution of the L 7 pairs Let f L H c denote the density of L H pairs under the null hypothesis that responses are independent and identically distributed as N 0 0 The exact form of f L H o will not be discussed here however one may show that to a good approximation we have f L H o fi L o H for a suitable density f That is the length of segments does not depend on the null variance and the height of segments scales as o Accordingly we need only find the distribution of L H pairs for independent and identically distributed N 0 1 responses in order to determine f L H o for all o 36 Suppose now that Z H is an approximation to j L H found by Monte Carlo simulation using N 0 1 responses For any given ec gt 0 we approximate the density f L H o by f L H o f Le H Figu2. generally most useful for the purpose of visualizing the data CGH Explorer implements two different smoothers for use in visualization of array CGH data the moving average smoother and the edge preserving smoother Whereas the former produces estimated log ratios f that typically vary from gene to gene although less than the original measured log ratios the latter produces estimated log ratios that stay constant over regions The moving average smoother As before we have a number of pairs x y i 1 m where x is the position of the ith gene x x and y is the corresponding log ratio Let w gt 0 the neighborhood size be a given integer Define f to be the running mean of the data y using a symmetric nearest neighborhood and 2w 1 neighbors Specifically for genes not close to the boundary k w 1 n k we define hi Usu RE Yk w 1 spem p Yr w w 1 For the remaning genes k 1 w and k 2 n w 1 n we let s max 1 k w and t min n k w and define i Ys ae ee bee 31 The edge preserving smoother As before we have a set of points x y i 1 m ordered such that x lt lt x An edge preserving smoother seeks to determine a sequence f that approximates the sequence y well and contains as few jumps as possible by a jump in the sequence f we mean a pair of consecutive elements f and f that satisfies f f The edge preserving smo
3. Search for genes with names containing a particular phrase Use to identify all points for which the gene name contains a word such as transcription by typing the phrase transcription without the apostrophes in the search dialog Use to identify all points for which the gene name is equal to a particular phrase such as EST by using the phrase EST with the double apostrophes Clear search result Restore the original plot The Preferences menu is used to change the appearance of the plot Menu entries are Chromosome separators Show hide vertical lines separating the chromosomes in plots that span several chromosomes Grid Show hide horizontal guiding lines positioned at the same height as the tick marks on the vertical axis Cytoband ID Show hide names of the cytobands When this feature is on the names appear at the bottom of the plot above the ideogram Tick marks Adjust the spacing between the tick marks on the vertical axis and the spacing between the grid lines if grid is turned on Grid type Choose between normal equispaced tick marks grid and exponentially spaced tick marks grid in which the grid lines are positioned at 1x 2x 4x Vertical range Adjust the vertical range of the plot Plot preferences Change the size shape and color of points and the type width and color of lines 22 Line plot moving average smoother Show a line plot of the data A line plot is a smooth represen
4. available through the Tools menu Datasets Show the data as a table of numerical values Tables Make an internal copy of a data set This may be useful if you want to apply transformations to a data set and still want the original data set to be available for analysis Plots Impute missing values Missing values are always imputed when you import the data into CGH Explorer However CGH Explorer keeps track of which CGH values have been imputed and allows you to perform imputation again at a later time 28 The following operations can be performed on a data set all these operations are available through the Tools menu Working directory Specify a new working directory i e a new directory to use as the starting point in all file dialogs Query definitions Modify the query definitions Plot resolution Choose between enabling and disabling the low resolution mode In low resolution mode graphs plot only a subset of the data points when there are many points in a plot this may save considerable time Genome selection Choose between different organisms This affects CGH Explorer s expectation about the number and size of chromosomes as well as the cytoband information given in plots Look and feel This potentially affects the appearance of certain dialogs and windows in CGH Explorer You may choose between the cross platform look and feel the default and the native look and feel For example using Windows t
5. E E G H I J K L M o iS ARRAY 1 ARRAY 2 ARRAY 3 ARRAY 4 ARRAY 5 ARRAY B ARRAY 7 ARRAY B clid chr nucleotide accession gene name 0 22 0 18 0 34 0 03 0 01 0 21 0 2 0 25 IMAGE 1 1 13850 W15460 ESTs 0 27 0 18 0 08 0 06 IMAGE 2 1 71204 AA431426 Novel human gene mapping to chor 0 07 0 61 0 01 0 06 0 18 0 04 0 2 0 46 IMAGE 3 1 167764 H39221 ESTs Weakly similar to Kelch mo 0 2 0 17 0 18 0 03 0 01 0 01 0 01 IMAGE 4 1 187455 AA458878 Homo sapiens agrin precursor mR 0 24 0 38 0 07 0 26 0 55 0 2 0 04 IMAGE 5 1 236684 AA406019 _ interferon stimulated protein 15 kD 0 15 0 01 0 23 0 01 0 03 0 01 0 04 IMAGE 6 1 526125 AAD24391 ESTs 0 3 0 04 0 25 0 34 0 6 0 32 0 14 0 23 IMAGE 7 1 891909 AA176164 KIAA1273 protein 0 14 0 29 0 32 0 34 0 14 IMAGE 8 1 986694 AA147499 hypothetical protein FLJ10024 0 1 0 2 0 27 0 16 0 2 0 07 D 11 IMAGE 13 1 1710318 T50675 caspase 7 apoptosis related cyste 0 1 0 07 0 23 0 51 0 21 0 58 0 23 0 1 IMAGE 14 1 1823834 AA487912 guanine nucleotide binding protein 0 42 0 4 0 12 0 28 0 17 0 65 0 44 0 43 IMAGE 15 1 1830258 AA433851 E74 like factor 3 ets domain trans 0 17 0 21 0 11 0 07 0 01 0 01 0 06 0 24 IMAGE 16 1 3256973 N50962 hypothetical protein FLJ20430 0 14 0 07 0 18 0 17 0 01 0 04 0 01 IMAGE 17 1 4019416 W71984 tumor necrosis factor receptor supe 0 06 0 04 0 22 0 32 0 2 0 37 0 08 0 06 IMAGE 18 1 4103194 AA448257 ESTs 0 12 0 06 0 03 0 15 0 1 0 1 0 06 0 12 IMAGE 19 1 4104207 AAD35123 cytosol
6. all arrays divided by the number of arrays divided by the number of genes multiplied by 100 For example with 10 arrays and 1000 features on the array a total of 1500 significant features in all arrays would give 15 altered genes e Altered arrays the number of arrays displaying a particular gene aberration averaged over all aberrant genes i e genes that are called significant for at least one array This is the total number of significant features in all arrays divided by the number of arrays divided by the number of aberrant genes multiplied by 100 No distinction is made here between type of aberration i e an aberration is either an amplification or a deletion e FDR the positive false discovery rate pFDR This is an estimate of the ratio between the expected number of false positives and the number of positives See Appendix B for details e The remaining columns show for each array the number of significant features 25 Step 2 selecting the number of significant features You select your preferred trade off between type I error and type II error by clicking on one of the rows in the ACE table A graph of the results for that cut off point will then be shown in a separate window You can choose between three different plot types the first one shown below is the default use the Preferences menu on the plot window to change to one of the others 1 frequencies ACE cgh2 FDR 0 0012 zu K Search Preference
7. graphical user interface e Missing value imputation array centering and various data transformations e Advanced graphical exploration of array CGH data e Identification of regions of amplification and deletion e File export facility for plots and tables CGH Explorer is available as a binary executable for Windows platforms Java source code is also available making it easy to run the software on other platforms this requires installation of the Java 2 platform or higher CGH Explorer is described in Lingjarde OC Baumbusch LO Liest l K Glad I and Bgrresen Dale AL 2005 CGH Explorer A program for analysis of CGH data Bioinformatics 21 821 822 Software and documentation is available at http www ifi uio no bioinf Papers CGH There are no restrictions on the use or distribution of the program We are happy to hear about your experience with the program Comments ideas and suggestions for future improvements are most welcome Please contact Dr Ole Christian Lingjerde Dept of Informatics P O Box 1080 Blindern N 0316 Oslo Norway E mail ole ifi uio no Installation in Windows To install the program in Windows go to the CGH Explorer web site http www ifi uio no bioinf Papers CGH Software and download the software When you run the installer program the software will be installed in the directory C Program Files CGH Explorer xx where xx is the version number and the program icon will be installed on the
8. hold the Ctrl key down while clicking on elements in the list to select two or more adjacent elements select the first then select the last while holding down the shift key Then click the button add gt in order to copy these elements to the field Import arrays Use remove gt in a similar manner to remove columns from your current selection 4 3 3 Optional operations to be performed during import The import dialog also contains five check boxes see table below for a description Note that if you don t specify that you want to impute missing values with array mean then CGH Explorer will simply replace missing values by zeros 17 When you have filled in the empty fields in the import dialog window and clicks OK the data will be imported The program also produces a new file on the same directory with the same name as the data file and ending with cgh This file contains the information you submitted in the import dialog window and the next time you open the data file all fields will be filled in but you still have the option to make changes resulting in an update of the cgh file A comment regarding the option to perform automatic position range adjustment in some cases such as for the sample data on the web site nucleotide positions may be inaccurate and even extend beyond the actual range for a given chromosome This may happen for example if nucleotide positions are obtained on the basis of an early draft o
9. in 0 L H f H gt Finally compute the variance of the responses of the genes that belong to L H pairs that fall in and use the result as the desired variance estimate 39 Dealing with memory problems You may experience memory problems in CGH Explorer if you import large amounts of data or if you perform many memory intensive operations ACE in particular is memory intensive Memory problems may in the worst case lead to a program crash unless you take precautions There are two important factors related to memory usage e The total amount of memory that is available to CGH Explorer This is determined at program launch time by the operating system and cannot be changed during a program session e The proportion of the available memory that have been used so far This proportion increases every time you import a new data set make a plot or perform an ACE analysis Note that closing a window in CGH Explorer e g a plot window or a data selection window does not free any memory since closing a window simply removes the window temporarily from sight Closed windows can always be reopened using the Window menu Use the toolbar to see if you are in danger of experiencing a memory problem The toolbar tells you how much memory is available for CGH Explorer and how much of that memory is used so far Eus Fa ME ji EB ES 27M Memory used so far Total amount of by CGH Explorer memo
10. the mouse while keeping the mouse button down When you release the mouse button the resulting rectangle see figure below will define a zoom region and a new plot window appears on top of the old one showing only the part of the genome that contains the genes inside the zoom region You may repeat the process as many times as you like by zooming further in on the new plots made in this way Here is an example E Scatter plot cghd Search Preferences The figure below shows the result of the above zoom operation Observe that pointing with the cursor at a dot in a scatter plot yields a yellow square containing the name of the corresponding gene clone Point at the ideogram at the bottom of the plot to get information about the corresponding nucleotide position E Scatter plot Search Preferences To obtain information about all clones in a rectangular area of the plot drag a square just as you did to zoom in but this time press the Ctrl key during the operation When you 10 release the mouse button information about the clones corresponding to the dots inside the square are shown in a table m Ratio Sample Accession No Gene name j 3 534 ARRAY 4 N89721 _ protein phosphatase 3 formerly 2B catalytic subunit gamma isoform calcineurin A gamma 3 1 ARRAY 4 4 H13424 Putative prostate cancer tumor suppressor 3 037 ARRAY 4 R828
11. these segments significant However the precise start and end points of a segment may require adjustment because of the way segments are defined using local averages Consider a significant segment i e a segment for which the pair L H belongs to Q Let z be the associated responses Define K min 16 4 and consider all possible subdivisions of the segment into three subsegments of length L L and L respectively where 0 lt L L lt K and L L L I3 For each subdivision characterized by the pair L L3 compute the sum of squares ba E Ia L5 tt D eme br A i 1 DL In 1 and find a subdivision Z Z that minimizes the above criterion The genes associated with the responses z i L 1 L L are then reported as genes with loss if the average of the responses is negative and as genes with gain if the average of the responses is positive Estimation of the variance parameter c We need to estimate the null variance o of the log copy number ratios ACE performs this task by first identifying in all the given DNA samples regions of DNA that are likely to be normal and then calculating the variance based on data from those regions Let be a pilot estimate of o obtained by computing the variance of the responses of the genes that belong to segments for which L 10 Let f L H f L H e and choose 0 such that roughly 50 of the total NM gene measurements belong to L H pairs that fall
12. 0 0 0 bloloiloleileileiejeieieileileieie bloloiloeleileieieijelieieileileieie bloloiloleileileiejeileieileileieie P ojojooooooooooooo Each row in the table corresponds to a particular gene The first four columns are the clone id the gene name the chromosome number and the nucleotide position of the gene By default the remaining columns one for each array show a numerical code indicating the status of each gene in each array 1 for deletion O for normal and 1 for amplification Red cells indicate array features that are classified as amplified while green cells indicate array features that are classified as deleted Use the menu command View Values in table to replace the 1 0 1 status value for each gene by the log copy number ratios The table include information about all genes in the data set significant as well as non significant ones By default all genes are highlighted in yellow To highlight only those genes that are significant in at least k samples use the menu command View Select genes You may later return to highlighting all genes by using the menu command View Select genes and clicking the Cancel button Step 4 saving the table to file Use the main menu command File Save as to save the table as a tab delimited text file Note that only the rows genes highlighted in yellow are saved to file 27 The following operations can be performed on a data set all these operations are
13. 1 An interesting extension of the method described here would be to take into account the actual physical distance between neighboring genes In that perspective the currently implemented method essentially assumes a uniform distribution of genes along the chromosome 35 We now return to the question of assigning each gene to one of two groups depending on whether it is a candidate for loss or a candidate for gain Define a binary classification of the genes based on the signs of the running mean terms with a small modification Let sign y where sign y 1 if y gt 0 and sign y 2 1 if y 0 We now let amp amp unless amp 64 amp amp 41 amp 42 in which case amp amp In words c is the sign of 7 unless all four neighbors y y yi v have the opposite sign in which case c equals the sign of these four neighbors This classification rule is basically a robust version of the rule that assigns a gene to one of the two groups based on the sign of a local average of the responses around the gene The binary classification induces a partitioning J UJ U U Jg of the gene indices 1 2 n into R gt 0 sets of consecutive indices such that for any k l J and C G for any ke J and l J 4 For example if amp Ce C then genes 1 6 form a segment and J 1 6 and if amp Cs 6 then genes 7 and 8 form a segment and J
14. 34 ESTs 3 024 ARRAY 4 AA284268 ESTs Weakly similar to A49656 estrogen responsive finger protein efp H sapiens 2 796 ARRAY4 748411 ESTs E e 2 762 ARRAY 4 R531 12 platelet derived growth factor receptor like 2 646 ARRAY4 AA679352 famenydichsenhata farnesyltransferase 1 EN 2 545 ARRAY 4 AA085676 KIAA0942 protein 247 ARRAY 4 N21407 activated RNA polymerase II transcription cofactor 4 2452 ARRAY4 AA487460 dihydropyrimidinase like 2 2412 ARRAY 4 N25097 BCL2 associated athanogene 4 2 392 ARRAY4 4028963 ataxin 2 related protein 2 302 ARRAY4 aA478279 l indoleamine pyrrole 2 3 dioxygenase 2271 ARRAY 4 H81104 RAN binding protein 16 2 258 ARRAY 4 R37165 l CCR4 NOT transcription complex subunit 7 2 255 ARRAY 4 IN31948 F37 Esophageal cancer related gene coding leucine zipper motif 2 237 ARRAY4 R72076 jneuregulin 1 2 227 ARRAY 4 W93086 ESTs 2 20418 ARRAY 4 T55870 solute carrier family 20 phosphate transporter member 2 2 211 ARRAY4 44043998 gonadotropin releasing hormone 1 leutinizing releasing hormone 2 JARRAY 4 JAAA47 797 plasminogen activator tissue s Point the mouse at any row in the table and click the left mouse button This initiates a search for more information about the clone in that row By default Internet Explorer will start up and perform a search on Entrez based on the clone id given in the third column of the table titled Accession No The default
15. 8 AAA1N7A3 inr finner nrntein 151 nH7 R7Y 7 xL i eghdata ia 7 Data files should have a title row followed by a row for each gene There should be one column for each array in the experiment as well as columns providing positional and other information for each gene The file should contain at least the following columns in arbitrary order o A column for each sample array to be analysed giving the copy number ratios or some transformation of the ratios such as the logarithm of the copy number ratios for all genes on the array Empty cells are treated as missing values and will be imputed during import see below for details o Aclone gene identifier e g an accession number It need not be unique The chromosome 1 2 X Y where the gene is located in normal DNA o The gene position Positions should be given as the number of nucleotides from the start of the chromosome o The gene name O Rows in the data file should be ordered increasingly with respect to chromosome number 1 2 X Y and within each chromosome increasingly with respect to gene position CGH Explorer does not check this make sure to do so yourself before importing a file 15 4 2 Missing values in input data Blank fields in the data file are regarded as missing values At import missing values are automatically filled in using one of these imputation methods Imputation with zeros mputation with array mean i e impute
16. CGH Explor Graphical exploration and sta ist analysis of array CGH data User Manual October 2005 The Bioinformatics Group Department of Department of Informatics Institute for C University of Oslo Norwegian Rad Norway Norway Contents 1 Getting started ii iiicsscaciieacncadiedceeisitenasesncestivdancadbecretuandedgans 3 2 Installation seosceiisiadanhninatanicaedaatdoaaduasareddecdededncmuniausatadartiaas 4 3 A quick Tutorials reesctnetetacensantaededsavinaheaswensetaadeualeniecnaebnabane 5 A ANE t l DAN erm S 14 5 Data import and export eeeeeeeees 15 6 Data frames iioi iai nis esack cecus Gu E Coria Ra Da oa Tea MR ianen 19 T ANG File merid ucciso anesan L DU POT EFES RE Ya D dead 20 8 The Graph maU uoasvixe deck cerni aui rcu Ra E onS e EC manaa 21 9 The Tools MON e 24 10 The Detection menu esee 25 11 The Window melli ceieni eio dip Ue dirDedukuwa Dubia oo GG re nd ada 28 12 The preferences menu eene 29 Appendix ue 30 Appendix B usetkcectoriiga s tutid ia oaa Desde QURR oae Ck Cl TEE EVO RM ik 33 Plor qum T T 40 CGH Explorer is a program for visualization and statistical analysis of microarray based comparative genomic hybridization array CGH data Some key features of the program e Available for most platforms including Windows Linux Mac e User friendly
17. ays Hence the total nuV mber of hypotheses is NM which typically is a very large number Suppose each test is performed at level o A and that the tests are independent this is only approximately true in ACE The expected number of false positives is then approximately equal to NMa A assuming that the number of true positives is relatively small compared to the total number of hypotheses In order to determine the appropriate level o A to use in the tests we follow Storey and Tibshirani 2003 and consider the positive false discovery rate pFDR defined as the conditional expectation of V R when R gt 0 where V denotes the number of false positives and R denotes the total number of positives rejected hypotheses The following estimate for pFDR is used in ACE NMo X where S as before denotes the number of genes that belong to segments for which L H is in Q The ratio NMa A S is thus approximately equal to the expected proportion of false positives among all positives 38 Step 6 Report genes The final step in ACE is to report a list of genes assessed at a particular significance level to have altered copy number In CGH Explorer the significance level is selected by the user from a list of possible levels on the basis of the corresponding estimate of the positive false discovery rate for each level For the chosen level we find as explained above the segments for which the pair Z H belongs to 0 and call the genes in
18. ays In the latter case the statistical analysis will be based on properties derived from the whole set of arrays Suppose there are data from N gt 1 arrays each with M genes clones Each array corresponds to a sample or individual For each array the input to the algorithm is a set of suitably transformed see below copy number ratios in the following referred to as responses Vis J 1 m k ligt where corresponds to the jth chromosome arm and k corresponds to the kth gene clone on the jth chromosome arm note that M n n The transformation should be chosen to make the distribution of the responses approximately normal A common choice is to let the responses be the logarithm in base 2 of the copy number ratios In the following we assume that the responses y have been centered so that the expected response for normal DNA is zero here and below we refer to DNA with normal copy numbers as normal DNA When this condition is not met approximate Copy numbers vary slightly even among healthy individuals in a population However it is not common to have array CGH data of normal DNA from all individuals in a study and most array CGH studies seem to ignore copy number polymorphisms 33 centering can often be achieved by subtracting from each response on an array the average of all responses on the array Consider the problem of identifying segments of DNA that deviates from normal DNA The approach used b
19. desktop and under All Programs on the Start menu To uninstall CGH Explorer use the Windows facility Add or Remove Programs located in the Control Panel Installation in other operating systems To install CGH Explorer in other operating systems you need to download and compile the source code available from the CGH Explorer web site To compile and run the software you also need the Java 2 standard edition development kit and runtime environment J2SE JDK JRE These are available from Sun s web site look for Desktop Java on the web site http java sun com j2se The following tutorial is designed to get you started quickly with CGH Explorer The main components of the program are shown and you learn how to import a data file perform simple preprocessing of the data plot the data and search for copy number alterations For a complete overview and more details on CGH Explorer please see later chapters Before you start on the tutorial download the sample data set from the CGH Explorer download central and save it under the name cghdata txt Step 1 Launching the program for the first time When you launch the program for the first time a file dialog appears Use the dialog to specify the working directory i e the directory you want to use as the default start location for file dialogs when you import or export files Your selection will be stored for later sessions as well Normally you would like to select the directo
20. e Value A CLASSPATH CiljavaleasyIOleasyIO jar C Java ComSpec C WINDOWS system32 cmd exe FP NO HOST C NO fsecoppgr 2005 08 25 fsecver 5 42 J Delete 4l 4 The bottom list of system variables consists of two columns the left column is the name of the system variable and the right column is the value of the variable Make sure that no variable with the name JETVMPROP is defined if it is you probably have another program installed that depends on this variable What to do in that case depends on the circumstances and is outside the scope of this manual Click on the button New below the list of system variables and fill in exactly as follows the number 400 means that we extend the memory available to CGH Explorer to a total of 400Mb you may want to use another number here depending on your needs and the amount of available memory on your computer New System Variable variable name JETVMPROP Variable value Djet gc heaplimit 400m 5 Click on OK in all the dialog windows to finish 42 REFERENCES Storey JD and Tibshirani R 2003 Statistical significance for genomewide studies PNAS 100 9440 9445 Winkler G and Liebscher V 2002 Smoothers for Discontinuous Signals J Nonpar Statist 14 203 222 43
21. ere a is the arithmetic average of z Mean center arrays This replaces the CGH values 2 z for an array by new values z a 2 a where a is the arithmetic average of z 24 This menu contains a single command Analysis of copy errors ACE Use this command to detect amplifications and deletions The analysis is always performed on a whole data set marking a selection of arrays and chromosomes in the data window has no effect Step 1 computing the ACE table The first step of the analysis is to select a data set by pointing the cursor at the title bar of the corresponding data window and clicking the left mouse button and performing the Analysis of copy errors ACE command on the Detection menu Note before doing this make sure that 1 the data are on log scale and 2 the data are centered such that the expected value for a normal gene is zero The latter can often be approximately achieved by array centering the log transformed data After a while a table called the ACE table appears Rows in the ACE table correspond to different trade offs between two conflicting interests keeping the number of false positives low low type I error and keeping the number of false negatives low low type II error The columns in the table are e Altered genes the number of called i e significant genes on an array averaged over all arrays This is the total number of significant features in
22. f the human genome In such cases you may want to perform automatic position adjustment by scaling the position values to fit the length of the chromosome exactly This will only give a crude approximation to the correction positions and results obtained after such rescaling of the positions of genes on a chromosome should be interpreted with that in mind 4 4 Saving data tables to file To save a table made in CGH Explorer as a tab delimited file you have two options e Select the table i e activate the window by pointing at it and clicking the left mouse button and choose File Save As e You may also save a table or part of it by selecting all or some elements in the table making a copy of the selection by typing Ctrl C and then pasting the result into an Excel sheet 4 5 Saving graphics to file Plots produced by CGH Explorer can be printed or saved To save a plot first make the graph window active You now have three options e Choose File Save As to save the plot as a Postscript file e Choose File Print Graph and select a printer that gives you the option to print to file Check the box Print to file and select Print The file with the ending prn is essentially a postscript file and should be treated as such by postscript printers If you have installed Adobe Acrobat PDFMaker you may also be able to print the document to Adobe PDF to produce a pdf file e Choose Tools Convert
23. hapter 5 for a detailed description of all entries in this dialog E import of file C 0le Chr DNRICGHDatasets cghdata txt EE ER mu E Columns in file Import arrays Clone ID Gene names Chromosome imageno a clid x name w chro chro 23 27 oF sth Z add gt Position From row To row rame nucl v b 6405 FA ae bad 3 lt remove i ABBAT S Dataset name optional ARRAY 4 cghd ARRAY 5 EE Une A B C D E F 6 Hi Automatic position range adjustment 1 imageno chro nuel lid name ARRAY 1 ARRAY 2 ARRAY 3 ARRAY 4 Ellie ANES 2 IMAGE 1 13850 W15450 ESTs 1212 0773 0885 1 169 za 3 IMAGE2 1 71204 AA4314 Novelh 4 15 0924 0548 177 L Mean center arrays 4 IMAGES 1 467764 H39221 lESTs 1 12 1237 os88 0 839 n 5 IMAGE 4 1 487455 AA4588 Homo 0 87 142 0876 1164 _ Impute missing values with array mean 5 IMAGES 1 236684 AA4OSO intefer 1 148 0 995 1281 106 7 IMAGES 1 526125 AA0243 ESTs 0815 oses 0764 1 176 d vj Remember settings to next session dE OK Cancel For our data set we need to specify which columns in the file are to be treated as CGH data Select the columns ARRAY 1 ARRAY 8 in the list titled Columns in file click on ARRAY 1 in the list then hold down the shift key while you click on ARRAY 8 in the list Use the bu
24. he native look and feel will result in Windows looking file dialogs Note that changes in the look and feel in the middle of a session may give some unexpected results As a rule change the look and feel immediately after starting a session or at the end of the session this will affect the next session On some computers a change in the look and feel may not result in any change of the appearance 29 Smoothing For any given array and chromosome we have a number of pairs x yj i 1 m where x is the position of the ith gene and y is the corresponding CGH measurement assumed here to be given as a log ratio Assume that the pairs have been ordered from left to right on the chromosome i e such that x x Each CGH measurement y is determined not only by the actual copy number ratio for that gene but also by a number of other factors including gene dependent and microarray dependent factors A simple but useful model for the CGH measurements is y f noise 1 1 2 m where f is the actual log copy number ratio for the ith gene and noise is the combined contribution of all other factors After appropriate normalization and centering of the data the noise terms noise i 1 m should be approximately independently and identically distributed Since our main interest is in the actual log ratios and not the measured log ratios we would like to estimate the two components f and noise from their su
25. he title bar of the data frame and clicking the left mouse button you may point anywhere else in the data frame when you click the mouse button but that will erase your current subset selection Apart from plot commands all operations on data sets are performed on the whole data set even if you have selected a subset If you do need to perform one of these operations on a subset of the data you have to prepare a new data file consisting of the relevant subset and import this file into a separate data frame Step 4 Log transforming the data It is often preferable to work with log transformed ratios rather than ratios For example the log transform tends to make the variation of the values less dependent on the magnitude of the values and the log transform also reduces the skewness of highly skewed distributions Amaratunga and Cabrera 2004 In addition some of the tools in CGH Explorer including the detection of deletions and amplifications assumes that the data are on logarithmic scale Log transforming the data is one of the optional actions you can take during data import see step 2 It can also be achieved after import as follows First make sure the data frame is active Transformations in CGH Explorer are always applied to the whole data set so it does not matter what the current selection of arrays and chromosomes are Give the command Tools Transform and pick the transform Log2 x Click OK to apply the transf
26. ic acyl coenzyme A thioest 0 16 0 23 0 1 0 71 0 2 0 26 0 03 0 09 IMAGE 20 1 4397122 H14383 potassium voltage gated channel 0 14 0 25 0 22 0 18 0 1 0 16 0 07 0 03 IMAGE 21 1 5477975 HO6156 Homo sapiens clone 23927 mRNA 0 54 0 25 0 32 0 32 0 36 0 04 0 25 IMAGE 26 1 9143144 N68404 Homo sapiens cDNA FLJ22807 fis 0 03 0 01 0 17 0 11 0 51 0 34 E 04 Io 08 IMAGE 27 1 9143144 R55630 Homo sapiens cDNA FLJ20895 fis 0 42 0 07 0 07 0 04 0 12 0 15 0 14 0 06 IMAGE 28 1 9820020 AA487452 DNA fragmentation factor 45 kD a 0 36 0 1 0 01 0 09 0 12 0 19 0 1 0 06 IMAGE 29 1 9981947 AA596759 phosphogluconate dehydrogenase 0 01 0 2 0 27 0 51 0 25 0 32 0 15 0 2 IMAGE 30 1 10014609 N50745 cortistatin 0 45 0 11 0 18 0 2 0 14 0 04 0 11 0 19 IMAGE 31 1 10142104 W38657 ubiquitination factor E4B homologe 0 07 0 25 0 12 0 1 0 1 0 22 0 03 0 18 IMAGE 32 1 11343540 N69283 TAR DNA binding protein 0 12 0 14 0 17 0 07 0 14 0 17 IMAGE 33 1 11355735 W88792 ESTs 0 04 0 4 0 04 0 16 0 14 0 06 0 12 IMAGE 34 1 11383802 AA669545 spermidine synthase 0 27 0 01 0 14 0 08 0 12 IMAGE 35 1 12174669 AA476240 procollagen lysine 2 oxoglutarate 0 22 0 15 0 29 0 28 0 38 0 04 0 03 IMAGE 40 1 15984733 AA281152 caspase 9 apoptosis related cyste IMAGE 41 1 15029917 H80637 KIAAD9B2 protein 0 27 0 04 0 25 0 04 0 12 0 04 0 15 0 15 IMAGE 42 1 16154822 H02333 ESTs Highly similar to regulatory nnz 0 36 nni n nsa na MAGE 43 1 1845042
27. ional operations to be performed on the data during import The various components of the import wizard dialog are now described 4 3 1 The preview table A condensed view of the data file a preview table is shown in the lower right corner of the import wizard Notice that some columns are shaded in green The shaded columns are those that are selected for import so far The preview table serves two purposes it tells you how CGH Explorer splits the file into columns this may be useful to detect errors in the file format and it tells you which columns have been selected for import 16 4 3 2 Column specification Notice that four columns have already been selected for import CGH Explorer has recognized the titles given at the top of these columns chro nucl clid and name and knows what function to assign to these columns For example a column with the title chro will be interpreted by CGH Explorer as a column with gene identifiers If you want to use another column as your gene identifier select another column in the selection box Clone ID in the dialog The table below gives an overview of entries in the dialog that are related to the selection of columns The first entry in the table Import arrays allows several columns to be selected whereas other entries allow only one column to be selected To specify which arrays to analyze select one or several columns in the left hand list to select two non adjacent elements
28. lorer are available from the menus Some of the most commonly used commands are also available on the tool bar Plot moving average fit as curve plot Perform analysis of copy errors ACE Total available memory for Plot moving program used average fit as and unused stem plot Plot density of copy numbers Save to file Heat map i oS grana t t t t Import Plot spatial Plot edge data file density of preserving fit genes as curve plot Memory use green used gray unused Scatter Plot edge plot preserving fit as stem plot To use the tool bar first select the target object data frame graph or table of the command For example to print a graph you first select the graph window by left clicking the mouse on the window and then left click on the print symbol on the tool bar To make a plot you have to 1 select a data set by left clicking on a data selection window 2 select the arrays and chromosomes that you want to plot and 3 left click on the wanted plot symbol on the tool bar 14 CGH Explorer allows you to import and analyze simultaneously as many data sets as you like limited only by your computer s memory and speed 4 1 Input data file format To analyze a data set consisting of any number of arrays first collect the data in a single tab delimited text file Here is an example of a data set shown in Excel Ls cghdata xls js x A B C D
29. m y f noise In order to do this we make an important assumption here stated informally that the sequence of actual log ratios f i 1 m fluctuates much slower than the corresponding sequence of measured log ratios y i L m This is reasonable to assume if segments of normal DNA for which f 2 0 and segments of altered DNA in which either f gt 0 or f O0 holds for all genes in the segment typically include several genes Much stronger assumptions than above are sometimes made about the data generating process This leads to various types of models such as Markov chain models and ANOVA models Precise estimates of the wanted quantities such as the f and their uncertainty may be obtained under such assumptions but the quality of the results may depend heavily on the validity of the model assumptions Although often very useful 30 such procedures should only be applied after careful consideration of how well the data meet the stated assumptions A useful alternative particularly for early visual inspection of the data is to apply a smoothing procedure Smoothing procedures make only weak assumption about the data generating process basically that the actual log ratios fluctuates slower than the measured log ratios thus they are flexible and less prone to model misspecification On the other hand smoothing procedures should not be expected to produce very precise estimates of the wanted quantities and they are
30. nt to see the vertical lines in the plot that indicate the boundaries between the chromosomes you may go into the Preferences menu on the graph window to turn this feature off At the bottom of the plot you see what chromosomes are shown as well as cytobands for each chromosome To demonstrate another way of visualizing the data reactivate the data frame by clicking on the title bar of the data frame and give the command Graph Stem plot Moving average smoother WWW Ql Den Gd a i ETC This is a stem plot that shows only a fit to the data and not the data themselves Rather than plotting the fit as a curve the fit is plotted as a sequence of vertical bars one for each gene The height of a bar is the fitted value for that gene The fitting procedure in this case is a moving average smoother see Appendix A for details Various plot options are available from the Preferences menu on the plot window You can hide the vertical lines separating the chromosomes superpose a grid of horizontal lines and show names of cytobands The vertical plot range the number of tick marks and the color and size of the points are also adjustable The Search menu on the plot window allows you to search for and identify all genes that have a certain phrase in their gene name Step 7 Exploring details in a plot Using the mouse you can zoom in on any part of a plot Just point the mouse inside the window click the left mouse button and move
31. oned along the chromosome We assume that each gene has been assigned a unique position which may for example be the position on the chromosome of the first nucleotide of the gene As a first step we want to assign each gene to one of two groups depending on whether it is a candidate for being part of a loss region or part of a gain region Note that at this stage of the analysis we do not seek to determine the strength of evidence in favor of loss or gain and accordingly there is no third group corresponding to those genes that are neither candidates for being part of a loss or a gain region In order to make the assignment of each gene to one of the two groups we consider the gene as well as a small neighborhood around the gene The method to be described here does not utilize the actual genomic distance between neighboring genes only the order of the genes along the chromosome Hence in referring to a small neighborhood we only intend to imply that it is small with respect to the number of genes in the neighborhood Let y denote the running mean of the data y using a symmetric nearest neighborhood and 2w 1 neighbors in the version of ACE now implemented in CGH Explorer we use w 2 For genes far away from the boundary k w lL n k we have Te Yk w Ucwaa os Jeu 2w 1 For genes close to the boundary k 1 w and k n w cl n we define s max L k w and t min n k w and let Vk Ys eios y t s
32. ormation to the data Step 5 Plotting the empirical distribution To see the empirical distribution of your data first select the arrays and the chromosomes in the data frame that you want to consider Then give the command Graph Density of CGH values A plot similar to this one will appear F Density cghd mug Preferences mean 1 182 var 0 479 skew 2 324 kurt 10 078 A density histogram is shown together with a Gaussian density fitted to the data using maximum likelihood The mean variance skewness and excess kurtosis of the data are shown in the upper right corner a normal distribution has zero skewness and zero excess kurtosis To alter the horizontal range of the plot or the number of bins in the histogram use the Preferences menu on the plot window Notice that there is some skewness 1 e a lack of symmetry in the distribution present and also some positive kurtosis i e heavier tails than for a normal distribution Step 6 Plotting the data Activate the data set and select a subset of the data e g chromosomes 7 to 16 and ARRAY 4 We now demonstrate two ways of visualizing the data The first is a scatter plot of the data along with a curve fitted to the data using an edge preserving smoother see Appendix A for details about smoothers To create this graph give the command Graph Line plot Edge preserving smoother L a O Se 3 3 10 2 93 4 15 If you don t wa
33. ot of the data A stem plot is a smooth Fue sae e representation of the data and is similar to a line plot NE except that smoothed values are represented as vertical bars red for positive values and green for negative values rather than points joined by line segments See ll MI I Appendix A for a description of the moving average smoother Use the Smoother menu on the graph window to alter the smoothness of the estimate See scatter plot for a description of the other menus on the graph window 23 These operations can be performed on a data set the first can also be applied to a graph Convert to table Convert a data set or a graph to a numerical table Duplicate Make an internal copy of a data set This may be useful if you want to apply transformations to a data set and still want the original data set to be available for analysis Impute Impute missing values Missing values are always imputed when you import the data into CGH Explorer However CGH Explorer keeps track of imputed values and allows you to reimpute at a later time for example after you have applied a transformation to the data Two imputation methods are available impute with zeros or impute with array means Transform Apply a transformation to all the CGH values in a data set The following transformations are available Mean center genes This replaces the CGH values z for a gene by new values z a n a wh
34. other that is implemented in CGH Explorer is known as a Potts filter see e g Winkler and Liebscher 2002 Given a penalty parameter 2 gt 0 the values T are found by minimization over the scalars f f of the penalized least squares criterion m Hf fd X O 4 ti i l Ja The first term on the right hand side is a goodness of fit term The last term on the right hand side is a regularization term that penalizes for jumps in the function values It equals a nonnegative constant A times the number of incides i for which f f Thus a constant function for which f f has zero penalty whereas a function that changes value in every point has maximal penalty The magnitude of A controls the trade off between goodness of fit and smoothness defined here as the number of jumps The solution to the above optimization problem can be found using dynamic programming we skip the details here 32 The ACE algorithm Here we discuss the technical details of the Analysis of Copy Errors ACE algorithm A more detailed account of ACE including further developments and refinements of the method will be submitted for publication in the near future The discussion below pertains only to the current implementation of ACE in CGH Explorer Basic concepts The purpose of ACE is to search for copy errors in array CGH data The ACE algorithm may be applied to a single array or simultaneously to a collection of arr
35. possibility of outliers may have to be ruled out first Second genes in segments of moderate length and low height the shaded areas in Figure 1 are not called significant even though the null density is low in this part of the L H space The reason for this is that we don t expect the density in this part of the L H space to be higher under the alternative hypothesis 37 that would essentially imply that the variance is smaller under the alternative hypothesis than under the null hypothesis and we have seen no empirical evidence in favor of this Define f L H max f L h h gt HY For gt 0 define the acceptance region Q G8 LE Lus V P LH 2 and the corresponding rejection region M IH L gt La P 5H A In the version of ACE currently implemented in CGH Explorer we have Lam 0 i e there are no lower limit on the length of the segment to which a significant gene belongs For a sequence of values 0 lt 4 lt Ap we compute the proportion of genes that belong to L H pairs that fall in under the null hypothesis We refer to this proportion as a significance level and denote it by a A We also compute the number of observed genes S that belong to L H points in Q we refer to this number as the number of significant genes for the given level A Step 5 Estimate the positive false discovery rate Recall that M is the total number of genes on the arrays and N is the number of arr
36. re 1 shows the approximation f L H resulting from one particular Monte Carlo simulation approximations converge reasonably fast as a function of the number of simulated L H pairs In practice the null variance has to be determined from the data Suppose c is an estimate of c see later section on how is defined Then our approximation for the 82 null distribution of the L H pairs will be f L H f L H a b A J 1 0 0 5 4 S S 00 AC 4 UC 10 20 30 L 10 20 30 T Figure 1 Contour plots showing the empirical distribution of L H pairs for simulated null data with variance c 1 a The border between the rejection region right region and acceptance region left region for one particular significance level a is shown as a broken line b The border between the rejection region right region and acceptance region left region for a different significance level a gt a is shown as a broken line Step 4 Find significant genes A gene is called significant in ACE if it belongs to a segment for which the density L H is below a given threshold This is analogous to how significance is defined in e g a common t test However in ACE there are two situations that are treated as exceptions from the above rule First genes in very short segments say L lt Lmin are not called significant even if the density is below the treshold since the
37. ry available to CGH Explorer 40 Suppose the proportion of memory used so far exceeds some threshold say 75 of available memory and you want to import new data or perform further analyses You may then want to consider terminating the current program session and increasing the total amount of memory available to CGH Explorer To do this you must define an environmental variable called JETVMPROP Below we explain how you do this in Windows XP 1 Open the Control Panel in Windows and start the program called System note that the Control Panel and or the program System may be hidden from view depending on your Windows preferences in that case you must change Folder Options 2 Click on the Advanced tab The result so far should be as follows System Properties System Restore Automatic Updates Remote General Computer Name Hardware Advanced You must be logged on as an Administrator to make most of these changes Performance Visual effects processor scheduling memory usage and virtual memory Settings User Profiles Desktop settings related to your logon Settings Startup and Recovery System startup system failure and debugging information Settings Environment Variables Error Reporting Cx 3 Now click on the button Environmental Variables to show a window like this Environment Variables User variables for ole Variable Value New Edit Delete System variables Variabl
38. ry where your CGH data files typically reside You can change the location of the working directory at any time using the Preferences menu Step 2 Importing data Having selected the working directory the application main window appears Detection Window Preferences Help LIFE RATER Notice the tool bar located below the menu bar Many commonly used commands are available from the tool bar however you should nevertheless familiarize yourself with the menus in order to make the best possible use of CGH Explorer We import the sample data file cghdata txt available from the CGH Explorer download central you need to download the file to your file system before continuing using the command File Import data also available leftmost on the tool bar Locate the data file and open it The import function recognizes tab delimited text files with a specific column format and with names that end with txt The data file format is described in chapter 5 If your data consist of a series of files one for each array create a new file with columns for gene ID gene name chromosome number position on chromosome measured in nucleotides and one column of CGH data for each array Then save the file as a tab delimited text file This is all easily done in e g Excel The import wizard now appears see figure below This dialog is used to specify how to interpret the columns in the data file See c
39. s percentage amplified and percentage deleted For each gene the percentage of arrays with amplification of that gene is shown in red and the percentage of arrays with deletion of that gene is shown in green 2 alterations single plot E ACE cgh2 FDR 0 0012 zu bi Search Preferences Each copy number alteration is shown as an interval red for amplifications and green for deletions There is one row for each array ordered from top to bottom 3 alterations split plot E ACE cgh2 FDR 0 0012 mu bi Search Preferences Copy number alterations Each copy number alteration is shown as an interval red for amplifications and green for deletions There 1s one row for each array ordered from top to bottom Amplifications and deletions are shown separately N w n a o e o 5 26 Use Preferences in order to change the appearance of the above plots e g to change the color scheme in the alterations plots above You may zoom in on a particular part of the plot as described earlier in the section on graphical exploration Step 3 converting to a numerical table The graph above can be converted to a tabular representation by making the graph window active and giving the menu command Tools Convert to table Here is an example of such a table ARRAY 1 ARRAY ZARRAY JARRAY JAR D D D 0 0 0 0 0 0 0 0 0 0 0
40. search operation mentioned above may not be appropriate for you in which case you can define your own search using Preferences Query definitions in the menu on the main window Step 8 Detecting copy number alterations CNAs In order to detect copy number alterations CNAs we select the Analysis of copy errors ACE command on the Detection menu After a while a table referred to as the ACE table appears F ACE cghd mu K Altered genes Altered arrays FDR ARRAY 1ARRAY 2ARRAY ARRAY 4ARRAY Before you apply ACE to you data always make sure that your data are on log scale and that the log ratios are centered such that the expected log ratio for a normal gene is approximately zero Otherwise the ACE algorithm fails to work properly 11 Each row in the ACE table corresponds to a particular trade off between type I error false positives and type II error false negatives Select any row by pointing the cursor at it and clicking the left mouse button A new window will then pop up similar to this one F ACE cghd FDR 0 0001 zug E Search Preferences percentage amplified and percentage deleted i doe lide 100 Coc ie EE E E E EZ LZ Oooo 1 4 5 6 7 8 9 10 11 12 13 14 1516171819202 For each gene the height reflects the proportion of arrays for which the gene is amplified or deleted Po
41. sitive values are shown in red and indicate the proportion of arrays for which a gene is classified as amplified Negative values are shown in green and indicate the proportion of arrays for which a particular gene is classified as deleted Using the Preferences menu on the graph window you may change the plot type to alterations split plot to show instead the individual copy number alterations CNAs in each array E ACE cghd FDR 0 0001 Search Preferences Copy number alterations 0a DL 8 9 10 11 12 13 14 15 1617181202 Here amplifications are shown in the upper part and deletions in the lower part of the plot Each row in the upper lower part of the plots corresponds to one array ordered from top to bottom 12 The above graph can be converted to a tabular representation by making the graph window active and selecting the menu command Tools Convert to table Here is an example E Copy number changes View Each row in the table corresponds to a particular gene The first four columns are the clone id the gene name the chromosome number and the position of the gene on the chromosome The table above can be exported to file The resulting file may also be imported back into CGH Explorer as a data set in order to plot loss and gain regions for individual samples See chapter 10 for more information 13 All commands in CGH Exp
42. so shows a normal density fitted by maximum likelihood to the subset of the data that is inside the interval a b above In the upper right corner of the plot the mean variance skewness and excess kurtosis of the data are shown for a standard normal distribution these have values 0 1 0 0 respectively Density of genes Plot and estimate of the spatial distribution of the genes ame ss using an Epanechnikov kernel density estimator Use the Density menu on the graph window to change the Y smoothness of the estimate See scatter plot for a rier description of the other menus on the graph window Some details on a chromosome let 2 be the gene positions and define a uniform grid In each grid point t the plotted value is proportional to ft Sor V5t z w where r t 30 t 5 v5 for t lt 5 and r t 0 otherwise The smoothness of the estimated curve is determined by the size w of the neighborhood By default w 4Mb and the total number of grid points on the genome is 5000 21 Heat map Plot a heat meap of the CGH values The rows in the Gteatmapcitoa z E heat map correspond to arrays ordered from top to bottom and the columns correspond to genes ordered from left to right Green squares correspond to negative values and red squares correspond to positive values Scatter plot Show a scatter plot point plot of the data The Search menu consists of two commands Find gene
43. t of such files into CGH Explorer The file dialog that appears during import always defaults to show the working directory To change the working directory use the Preferences menu Save As Save a table or a graph to file Tables are saved as tab delimited text files whereas graphs are saved as postscript files The file dialog that appears always defaults to show the working directory To change the working directory use the Preferences menu Page Setup Set print preferences Print Preview Show a print preview on screen Print Graph Print a graph Exit Exit the program Note that all graphs and other results are lost unless you save them before you exit 20 Use the graph menu to visualize a subset of your data First make the appropriate data window active by pointing the mouse at the title bar of the data window and clicking the left mouse button Then select the desired subset of arrays and chromosomes in the data window Finally select one of the commands on the graph menu most of them are also available on the tool bar 6 1 Description of commands Density of CGH values Plot the distribution of the CGH values as a histogram The default horizontal range of the plot is a b where a is the 0 5 quantile of the data and b is the 99 5 quantile of the data Using the Preferences menu on the graph window you can change the horizontal range and the number of classes bins in the histogram The graph al
44. tation of the data It is useful for visual determination of systematic alterations in the magnitude of the copy number ratios See Appendix A for a description of the moving average smoother Use the Smoother menu on the graph window to alter the smoothness of the estimate See scatter plot for a description of the other menus on the graph window Line plot edge preserving smoother Show a line plot of the data A line plot is a smooth representation of the data It is useful for visual determination of systematic alterations in the magnitude of the copy number ratios See Appendix A for a description of the edge preserving smoother Use the Smoother menu on the graph window to alter the smoothness of the estimate See scatter plot for a description of the other menus on the graph window Stem plot moving average smoother Show a stem plot of the data A stem plot is a smooth D 3 5 representation of the data and is similar to a line plot except that smoothed values are represented as vertical bars red for positive values and green for negative values rather than points joined by line segments See la l Ji Ili il Appendix A for a description of the moving average mq n smoother Use the Smoother menu on the graph window m mmm ms re mus to alter the smoothness of the estimate See scatter plot for a description of the other menus on the graph window Stem plot edge preserving smoother Show a stem pl
45. to table to convert the graph to a numerical representation in a table This table may then saved as described in the preceding section 18 An imported data set is initially shown as an empty table with one column for each chromosome and one row for each array referred to as a data frame A data frame serves two purposes e All operations on a data set will be performed on the currently active data frame To make a data frame active point at the title bar and click the left mouse button e Prior to some operations such as making a plot of the data the user is required to specify what part of a data set to use This is done by using the mouse to select one or several of the boxes each corresponding to a particular chromosome and a particular array in a data frame In the current version of the program selection of multiple boxes in a data frame is restricted to selecting a rectangular region in the data frame You may temporarily close a data frame by left clicking on the symbol in the upper right corner of the window The data set will still be available in CGH Explorer use the Window menu to reopen the window 19 Import data Import a data file Data files are text files consisting of copy number data and other information for a number of genes or clones and for one or more arrays See chapter 5 for a description of the structure of data files and the various options related to impor
46. tton add gt in order to copy these list elements to the field titled Import arrays There are some actions that you commonly want to perform on the data at import such as log transforming or mean centering the CGH measurements See the check boxes in the import wizard for such optional actions These actions are performed in the order that they appear in the dialog from top to bottom In this tutorial we do not change the settings for the optional actions Click OK to proceed Step 3 The data frame The imported data set is shown in CGH Explorer as a data frame see figure below Rows in the data frame correspond to arrays and columns correspond to chromosomes E Data cghd Plot actions in CGH Explorer are always performed on subsets of a data set To select a subset mark a rectangular area in the data frame using the mouse Below we have selected the subset consisting of the first four arrays and chromosomes 7 to 16 F Data cghd 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 H HA Immediately after import of a data file the newly created data frame is active you can see this by noting that the color of the title bar on the data frame is blue Some actions such as the creation of a plot will render the data frame inactive You then need to reactivate the data frame if you want to perform any further operations on the data set You do this by pointing the cursor at t
47. with the average of all values on an array 4 3 The import wizard Suppose you have downloaded the sample data set cghdata txt from the program s web site Choose File Import Data pick the file cghdata txt and click OK The import wizard will then appear F import of file C 0le Chr DNRICGHiDatasets cghdata txt i 27 eee dep Bd Columns in file Import arrays Clone ID Gene names Chromosome a clid k name hd chro hd add gt Position From row To row nucl v 2 6405 lt remove Dataset name optional eghd hs RA F A B C D E F G H l Automatic position range adjustment 1 limageno chre nuel elid name ARRAY 1 ARRAY 2 ARRAY 3 ARRAY 4 7 J zi _ Log2 transform values 2 IMAGE 4 19850 W15480 ESTs 1212 0773 oses 1169 E 3 IMAGE 2 1 71204 AAG 14 Novel h 1 16 0 924 0 648 1 77 _ Mean center arrays 4 IMAGE 3 4 187764 H39221 ESTs 1 12 1237 0 888 0 830 r3 XE E 5 IMAGE 4 1 187455 AA4588 H 0 87 1 12 0 976 1 164 _ Impute missing values with array mean SHE 6 IMAGE 5 1 236684 AA4060 interfer 1 146 0 995 1281 106 7 IMAGES 4 526125 AA02 02915 osses 0764 1175 vj Remember settings to next session 4 EZ Zu i OK Cancel This dialog is used to specify e which columns in the file to import and how to interpret them e opt
48. y true because of e g spatial effects on an array but that will be ignored here The 2 unknown variance parameter o is estimated from the DNA samples in the study using only regions that are likely to be normal details are provided further down ACE performs a series of tests to detect gene events a gene event being that a particular gene in a particular sample individual is subject to loss or gain There are NM null hypotheses each stating that a particular gene in a particular sample has normal copy If centering has not been performed on the data prior to import into CGH Explorer you may perform it in CGH Explorer by selecting the data set and choosing Data Mean center arrays ACE can easily be extended to utilize data from normal DNA samples when such data are available However this is not implemented in the current version of CGH Explorer 34 number hence follows a N 0 0 distribution The steps involved in the ACE algorithm are described next Step 1 Segmentation The purpose of this step is to identify in each array all possible candidate loss and gain regions Specifically we seek a subdivision of the genome into a minimal number of regions each of which are dominantly negative or dominantly positive with respect to the responses This will be made precise below Consider one array and let y y be the responses for all genes on a chromosome arm listed in the same order as the genes are positi
49. y ACE is first to identify all potentially interesting segments defined here as genomic regions for which the responses are dominantly positive or dominantly negative Properties of each segment their length and height see below are then compared with those of segments derived from normal DNA Clearly we need some knowledge of normal DNA to perform a comparison like this Such knowledge may be derived from several sources including normal DNA from the same individual and normal DNA from another individual or several other individuals These alternatives are useful when data for normal DNA samples are available but they cannot be applied when such data are unavailable In ACE on the other hand only data from the DNA samples that are included in the study are required for the analysis Hence ACE may be used even in situations where normal DNA is unavailable The formal statistical computations in ACE are based on the assumption that normal DNA responses are independent and identically distributed as N 0 c Note however that violations of this assumption mainly affects the computation of the positive false discovery rate and not the classification rule applied to the data to distinguish aberrant DNA from normal DNA The above assumption reflects the belief that response variability in normal DNA is due to the measurement process and the assumption that noise contributions for different genes should be independent this may not be exactl
Download Pdf Manuals
Related Search
CGH Explorer cg explorer