Pdf - Genome Biology
Contents
1. type non aligning multiply aligning paired end uniquely aligning or single end uniquely aligning Fig ure 2a The user can also run this tool looking at only the uniquely aligning reads to see if they align to exons introns intergenic regions or junctions Figure 2b The correlation tool generates either a heatmap or a a All read types o Junction Genomic Mult Nonmatch Millions of reads ivet_8 E SkelMuscie1_6 Brain A BET 1 Brain2_B SKelMuscle2_A Braini B Liveri A SkelMuscle1_A Liver2_A Liver2_B Braint A Liveri Liver _B Liver 5 Liver C Liver A Liver A Shalldueda A Shelldueca 8 Skealluscel A Brain2_6 Braint _A Brain 8 Brama _A SkalMusadsi A Skhallduede A Page 4 of 11 hierarchical clustering dendrogram showing the pairwise correlations of gene expression profiles in the RNA Seq or microarray samples of your project Figure 2c Sup plementary Methods in Additional file 1 For paired end data sets the pairdist tool shows the fraction of paired end reads for which 1 the two ends align to different chromosomes 2 the two ends align to the same chromosome but on the same strand 3 the two ends align to the same chromosome and differ ent strands but the minus end strand is upstream of the b Matching read types 80 60 40 20 Jateen ntron Percentage of matching reads SkelMusclez2 A Live r2_A Brain B Liver _B SkelMu2. Genome Biology ExpressionPlot x 7 ExpressionPlot a web based framework for analysis of RNA Seq and microarray gene expression data Friedman and Maniatis C BioMed Central Friedman and Maniatis Genome Biology 2011 12 R69 http genomebiology com 2011 12 7 R69 28 July 2011 Friedman and Maniatis Genome Biology 2011 12 R69 http genomebiology com 201 1 12 7 R69 Genome Biology SOFTWARE Open Access ExpressionPlot a web based framework for analysis of RNA Seq and microarray gene expression data Brad A Friedman and Tom Maniatis Abstract RNA Seg and microarray platforms have emerged as important tools for detecting changes in gene expression and RNA processing in biological samples We present ExpressionPlot a software package consisting of a default back end which prepares raw sequencing or Affymetrix microarray data and a web based front end which offers a biologically centered interface to browse visualize and compare different data sets Download and installation instructions a user s Manual discussion group and a prototype are available at http expressionplot com Rationale RNA Seg has emerged in recent years as the eminent platform for analysis of gene expression and RNA pro cessing 1 3 However processing the raw sequence data to get useful and accurate information about gene expression and RNA processing is still a daunting task even for computationally inclined researchers Hi
3. tergram where each point corresponds to a gene or event Here the x axis shows the change in that gene event in the first comparison and the y axis shows the change in the second comparison Figure 4 For exam ple points in the upper right quadrant would corre spond to genes events increased in both experiments whereas those in the upper left quadrant would be b 4 way Comparison Plot 01 0 0001 FC cutoff 2 P 2x2 2 39e 127 LOR 2x2 8 93 co estes utoft 0 E 46 0 360 a hg18 gene BurgeLab Human Tissue Panel RNA Seq 16 0 40x N C 40x 16 0 640X ey_A kidney_B kidney_C pancreas_A pancreas_B pancreas_CHiwer_A iwe Affymetrix Human Tissue Panel hg18 genes MAQE human MAQC_UHRi liver skel_muscle colon adipose 4 way Comparison Plot P cutoff 0 01 FC cutoff 2 P 2x2 5 45e 30 LOR 2x2 11 6 cor 0 C testicle A 122 0 441 mm genes C thymus_A thymus_B thymus 1 1 1 om m a a a an Ba H t an a a a t T 104 0 32 4 0X N C 40x 16 0 64 0X ey_Atkidney_B kidney_C pancreas_A pancreas_B pancreas_C liver_A live Affymetrix Human Tissue Panel hg16 genes T a wi a E a a a wi a a m Figure 4 Screen shots of ExpressionPlot 4way plots showing cross platform and cross species comparisons a Heart enriched gene expression in human tissue panel exon array 28 x axis and RNA Seq 1 y axis data sets Points c
4. of expression controlled background sets of similarly expressed but unchanged genes in terms of either RPKM or raw read numbers the user chooses although we recommend raw read numbers to avoid transcript length biases 21 These background sets are appropriate for downstream gene ontology or motif analysis A convenient feature of the table browser is the ability to click on any row to be presented with a link to the ExpressionPlot genome browser seqview This browser A Page 6 of 11 displays both RNA Seq reads including those spanning junctions as well as array probe intensities along with gene annotations described below Comparison of changes from different experiments data sets Having examined changes in two different conditions of a single experiment it is natural to ask how these changes compare to another experiment Sometimes this second experiment may be part of the same project but in other cases it could be part of another project and maybe even have been performed on another plat form for example RNA Seq versus microarray or in another organism for example human versus mouse The 4way tool and its associated table browser automa tically match up changed genes or RNA processing events from different experiments presenting them in a similar manner to its 2way cousin After selecting two projects and a pairwise comparison P value and fold change cutoff for each ExpressionPlot generates a scat
5. 2006 38 500 501 13 Fujita PA Rhead B Zweig AS Hinrichs AS Karolchik D Cline MS Goldman M Barber GP Clawson H Coelho A Diekhans M Dreszer TR Giardine BM Harte RA Hillman Jackson J Hsu F Kirkup V Kuhn RM Learned K Li CH Meyer LR Pohl A Raney BJ Rosenbloom KR Smith KE Haussler D Kent WJ The UCSC Genome Browser database update 2011 Nucleic Acids Res 2011 39 Database D876 882 14 Hubbard TJP Aken BL Ayling S Ballester B Beal K Bragin E Brent S Chen Y Clapham P Clarke L Coates G Fairley S Fitzgerald S Fernandez Banet J Gordon L Graf S Haider S Hammond M Holland R Howe K Jenkinson A Johnson N Kahari A Keefe D Keenan S Kinsella R Kokocinski F Kulesha E Lawson D Longden et al Ensembl 2009 Nucleic Acids Res 2009 37 D690 697 15 Anders S Huber W Differential expression analysis for sequence count data Genome Biol 2010 11 R106 16 Marioni JC Mason CE Mane SM Stephens M Gilad Y RNA seq an assessment of technical reproducibility and comparison with gene expression arrays Genome Res 2008 18 1509 1517 17 Robinson MD Oshlack A A scaling normalization method for differential expression analysis of RNA seq data Genome Biol 2010 11 R25 18 Bullard J Purdom E Hansen K Dudoit S Evaluation of statistical methods for normalization and differential expression in mRNA Seq experiments BMC Bioinformatics 2010 11 94 19 Affymetrix Affymetrix Power Tools http www af
6. during the website experience for easy navigation The manual link opens the page of the User s Guide relevant to the currently selected tool ExpressionPlot F xpressionPloty Web Server Zway 4way account correlation ecdf event heatmap geneley heatmap manual pairdist pairplot read types seqview 2way Plotter Away Plotter User Configuration Area correlation Pairwise correlations of transcriptional profiles ecdf Empirical Cumulative Distribution Function Plotter event _heatmap Heatmap levels or changes across multiple events and projects genelev Gene Level Barplotter he atmap Manual Compare change profiles from different sets User s Guide Paired end summaries Paired end Plotter Read types quantify aligning exonic intron etc Readmap Browser Figure 1 The ExpressionPlot home page The website opens with this screen giving a list of tools available in ExpressionPlot and a login box in the top right The navigation bar on top appears on all pages giving links to the other tools The manual link is context aware it automatically opens the User s Guide in another tab to the page explaining the current tool Friedman and Maniatis Genome Biology 2011 12 R69 http genomebiology com 201 1 12 7 R69 Quality control The ExpressionPlot front end provides several quality control tools for RNA Seq data The read_types tool graphs the number of reads in each sample of each
7. microarrays Background subtraction and probe normalization ExpressionPlot uses Affymetrix Power Tools 19 to perform the background subtraction using either Page 3 of 11 mismatch probes 3 UTR arrays or GC control probes exon arrays and follows this with quantile normalization of background subtracted probe intensi ties Users can use any affymetrix array for which they have the appropriate library files but for the following arrays those files can be automatically downloaded and installed by EP manage pl HG U133 A B HG U133_Plus_2 HuExon MOE430 A B MoExon and Rat230_2 Statistical calculations For microarray data gene levels are estimated first by finding all detected probes which are defined as probes with positive background subtracted intensities across all arrays in the project Once these probes are defined the gene level in each array is summarized as the med ian probe intensity P values for gene level changes are calculated by default using the Limma package 20 or optionally the t test As with the RNA Seq pipeline the P values are not by corrected for multiple testing unless specifically requested Web based front end global tasks Website users are initially presented with a landing page with links and short descriptions of all the different tools available in ExpressionPlot Figure 1 The naviga tion bar at the top as well as the login box on the top right are present on every page
8. or 3 of PACS exons Such events are created for all but the 5 most TSS and 3 most PACS Figure S6 in Additional file 1 Support for other types of events including alternative splice sites and sequence variants due to single nucleo tide polymorphisms or RNA editing is planned for a future release Statistical calculations For changes in gene expression ExpressionPlot uses the DESeq package 15 to model biological variation in the calculation of P values This package normalizes sam ples using median fold change and models the read counts using the negative binomial distribution includ ing a term for both sampling and biological noise Alternatively users can choose a modification of a pre viously described procedure 16 to detect technical dif ferences between two lanes or groups of lanes In a similar spirit to DESeq and other existing packages 17 18 total read counts are normalized using a robust procedure that is not dominated by the mostly highly expressed genes In this step the effective total number of reads in each sample is optimized to minimize the resultant number of significantly changed genes a pro cedure we call minimize significant changes Methods Supplementary Methods in Additional file 1 and exam ple data in Additional file 2 Finally a binomial test is performed on the number of reads aligning to a particu lar gene from the two samples to determine if the ratio is significantly differ
9. ample The two tailed P value is then calculated using R s binom test function Minimize significant changes method to estimate effective total read numbers To estimate the effective total number of reads n and n in a pair of samples or pair of groups of samples we estimate q2 which is the fraction of reads in the sec ond sample and then set ny qN and n N no where N is the total number of uniquely aligning reads from either sample The theory of our calculation of qz is that once a P value cutoff is set any potential choice of q will lead to a certain number of significantly changed genes say C qz which could be calculated by applying the procedure described above to every gene for example 27 389 genes in mouse Thus we have the optimization problem Friedman and Maniatis Genome Biology 2011 12 R69 http genomebiology com 201 1 12 7 R69 a MyD88 gene levels Page 9 of 11 b spleen enriched genes Gene level in set Affy HuExon Tissue Panel 500 400 z 300 m E 5 200 p 1h Khana ne li ep A a er A r 0 d er a er a o et a e O i er a a eraa a t a d e EEEEtTt gt saa E ogg agaaoccecingttg FEEFEE EE EEFE EEEE SSSAAA ssa Bee eo oe Tee EE i a ECDFs of Gene Set in Affy HuExon Tissue Panel 1 0 haari garabellum 0 2 0 4 0 6 0 8 Fraction of 316 Genes 316 total genes 0 0 Log2 Fold Change c spleen esas musca livar pancre
10. amples of a project or fold changes in the pairwise comparisons of a project Figure 6b Instead of looking at the distribution of the whole set the event_heatmap tool visualizes the individual levels or fold change of all the genes in the set as a heatmap Figure 6c Administrative tasks ExpressionPlot has an access management system that makes it easy for end users to share their data or release it publicly New user accounts can be made automati cally through the website including an e mail based password recovery feature When invoking the back end for a given project one user is assigned admin privi leges Users can then assign either view or admin pri vileges to other users on projects for which they are admin or can add a public flag to the project to make it visible without login These permissions are all con trolled via a simple web interface Download installation help Visit the ExpressionPlot website at 22 for instructions on how to download and install the latest version ExpressionPlot requires an existing MySQL and Apache web server as well as the RApache module The install pl script checks all the dependencies and tries to satisfy or make suggestions on how to satisfy any that are Page 8 of 11 missing It then downloads and installs the latest version of ExpressionPlot Alternatively a VirtualBox hard drive is available running Ubuntu linux with ExpressionPlot already install
11. as kidney fhyraid prostate braast caraballum haart WISE Kom ON ONKOL oe ot ee pe ed pen fa acm fra Ti iL me T m Osfoass gt EE ERETTE T EENE ETRE EEE EErEE GE ET oe tro gtoeg et 2 OF aoe A0 eons uo oun i z4 Un a Sey wi w Hz oO a 0 wi oF tas T pm pag 5 cann tt ie z q I Ea p u o u a 3 E gh 3 L g E T a Le Log2 Fold Change ELL Th LAS AP11 128616 1 004 Figure 6 ExpressionPlot screen shots examining spleen enriched genes in human exon array tissue panel data 28 a Levels of Myd88 a key signaling protein in the innate immune system 29 in human tissues using the genelev tool b ecdf showing tissue enrichment fold change relative to all other tissues of the 316 genes least 5 fold enriched in the spleen at a P value cutoff of 10 The sharp angle at 2 3 in the spleen curve indicates the 5 fold cutoff The position of the cerebellum curve to the left of all the others may reflect the general depletion of immune cells which is characteristic of the spleen within the nervous system c event_heatmap showing the fold enrichments of the 316 spleen enriched genes in all 11 tissues in the panel The screen shot was edited by removing many of the genes from the middle for formatting purposes and adding an arrow to indicate Myd88 which is part of a cluster of spleen enriched genes also enriched in the liver The depletion of the spleen enriched genes in the cerebellum is evident by the excess blue colo
12. berg SL TopHat discovering splice junctions with RNA Seq Bioinformatics 2009 25 1105 1110 5 Trapnell C Williams BA Pertea G Mortazavi A Kwan G van Baren MJ Salzberg SL Wold BJ Pachter L Transcript assembly and quantification by RNA Seq reveals unannotated transcripts and isoform switching during cell differentiation Nat Biotechnol 2010 28 511 515 6 Li H Ruan J Durbin R Mapping short DNA sequencing reads and calling variants using mapping quality scores Genome Res 2008 18 1851 18581 7 Katz Y Wang ET Airoldi EM Burge CB Analysis and design of RNA sequencing experiments for identifying isoform regulation Nat Methods 2010 7 1009 1015 8 Robinson JT Thorvaldsdottir H Winckler W Guttman M Lander ES Getz G Mesirov JP Integrative Genomics Viewer Nat Biotechnol 2011 29 24 26 9 Langmead B Trapnell C Pop M Salzberg SL Ultrafast and memory efficient alignment of short DNA sequences to the human genome Genome Biol 2009 10 R25 10 Wu Z Jenkins B Rynearson T Dyhrman S Saito M Mercier M Whitney L Empirical bayes analysis of sequencing based transcriptional profiling without replicates BMC Bioinformatics 2010 11 564 11 Goecks J Nekrutenko A Taylor J Galaxy a comprehensive approach for supporting accessible reproducible and transparent computational research in the life sciences Genome Biol 2010 11 R86 12 Reich M Liefeld T Gould J Lerner J Tamayo P Mesirov JP GenePattern 2 0 Nat Genet
13. cess article distributed under the terms of the Creative Commons Attribution License http creativecommons org licenses by 2 0 which permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited Friedman and Maniatis Genome Biology 2011 12 R69 http genomebiology com 201 1 12 7 R69 Tasks of gene expression analysis RNA Seq and microarray analyses begin with the follow ing pre processing tasks Back end pre processing tasks RNA Seq 1 alignment 2 read accumulation 3 statis tical calculations Back end pre processing tasks micro arrays 1 background subtraction 2 probe normalization 3 probe accumulation 4 statistical calculations The pre processing tasks are sequential and usually performed for all analysis projects In ExpressionPlot they are performed by the back end which is started from the command line on the server A typical RNA Seq data set might take a few days to run most of which is spent on alignments Using pre aligned data sets is possible by importing from BAM files Once the pre processing tasks have been completed the subsequent tasks can be considered a mixture of global discovery based and specific hypothesis based tasks In ExpressionPlot these tasks are the domain of the web based front end and all run on demand within seconds Global tasks a quality control b generation of plots and tables of changed genes events c genome w
14. eart skel_muscle colon adipo RPKM Up in brain 144 Incl Skip Ratio kel_muscle colon adipose testestlymph_node BT474 152 1e 04 1e 01 1e 02 1e 05 1B435 T47D MAQC_human MAQC_UHR liver heart skel_muscle colon adipo Incl Skip Ratio c Platform RNASeg P 1e 4 FC 1 5 Showing 144 144 Up skipped_exons Cao 1 36199264 36190317 Tens 1 33518208 33518249 e ft 118031780 118031812 chri 70282417 70282583 Human Tissues brain Up skipped_exons emo fa 123102155 123102225 170434 o PKM2 Convert hg18 skipped_exons to Gamma m Make Background controlling for Choose field V Go A Open Readmap Browser a Human Tissues e CLTA e chr9 36199264 36199317 n44 Figure 3 Screen shots of ExpressionPlot 2way plot and table_browser a 2way plot of human tissue panel RNA Seq data 1 showing brain gene expression on the y axis and average expression in all other tissues pooled on the x axis Blue points correspond to genes significantly higher P lt 10 fold change 20 370 points in brain relative to the other tissues green points correspond to significantly lower b 2way plot showing cassette exon usage inclusion skip read ratios instead of gene levels in the same data set The heavy lobe above the diagonal corresponds to exons with zero skipping reads in the brain and the lighter lobe below the diagonal corresponds to exons with zero skipping reads in all other tissues Although the P va
15. ed In either case after installation is com plete the EP manage pl script can be used to download and add on bowtie indexes annotations and microarray library files as required Example data sets both unpro cessed and processed can also be installed using the same script The User s Guide can be found at 23 and contains detailed instructions on setting up and running ExpressionPlot Please use the ExpressionPlot discussion group to post technical questions or hints This can be accessed by visiting the ExpressionPlot Google group 24 or by sending e mail to expressionplot googlegroups com Extracting biological meaning from high throughput data ExpressionPlot offers the gene expression community an easy to use tool for automated analysis of gene expres sion and RNA processing data The back end offers a solution to the problem of detecting significant changes in gene expression and RNA processing while the web based interface offers data analysis visualization and browsing tools that realize the biological potential of this new technology Methods Calculating P values for significance of changes in gene expression Given total numbers of reads in two samples or two groups of samples n and n2 g and go of which align to a particular gene of interest we model g as a bino mial distribution with parameters q3 and g where q2 nl n Nz and g g go is the total number of reads aligning to the gene in either s
16. end but there is at least one intron between the two ends The fifth category of reads where the two ends do not flank any known intron can be used to estimate the insert size and ECDFs of the insert sizes defined as the length of the un sequenced part of the library between the paired ends for the different lanes are also plotted by this tool Figure 2d data in Addi tional file 3 Generation of plots and tables of changed genes events The 2way tool and its associated table browser are the basic tools to examine the relationships between gene a gene level changes Page 5 of 11 levels or RNA processing events in two different samples The x axis will correspond to one sample such as wild type and the y axis to another such as mutant The project and pair of samples are chosen by the user from drop down menus and the plots like all the other plots in ExpressionPlot are generated on demand by the web server The 2way plot is a scatter gram where points correspond to genes or RNA pro cessing events for example cassette exons and are colored according to whether they are significantly dif ferent in the two samples Figure 3a b P value and fold change cutoffs for significance can be controlled by the user b cassette exon splicing changes Up in brain 370 muscle colon adinose tesies lymph_node BT474 1874 D 1e 04 1e 02 1e 00 1e 02 1e 04 1B435 T47D MAQC_human MAQC_UHR liver h
17. ent from the ratio of total numbers Friedman and Maniatis Genome Biology 2011 12 R69 http genomebiology com 201 1 12 7 R69 of reads in the two samples Supplemental Methods in Additional file 1 For the RNA processing events we form two by two contingency tables looking at the numbers of reads sup porting the two isoforms in the different samples for example Figures S3 S4 and S6 and Supplementary Methods in Additional file 1 The P values are then derived from either Fisher s exact test which is known to be conservative in this regime Supplementary Meth ods in Additional file 1 or if all the expected values are greater than 5 the Chi squared test By default the ExpressionPlot back end generates P values that are not adjusted for multiple testing This should be kept in mind when setting cutoffs on the web site We usually use a P value cutoff of 10 For exam ple using the UCSC genes cluster for mouse mm9 there are 27 389 genes so on average this cutoff would yield no more than 3 false positives Actually in most RNA Seq data sets many of the genes are not expressed or are expressed at extremely low levels and so the expected number of false positives is even lower since the small P values are not achievable for these genes Users who prefer to work with Benjamini Hochberg corrected P values can choose to do so by providing the correct switches as described in the User s Guide Pre processing tasks
18. fymetrix com partners_programs programs developer tools powertools affx 20 Smyth GK Linear models and empirical bayes methods for assessing differential expression in microarray experiments Stat Appl Genet Mol Biol 2004 3 Article3 21 Oshlack A Wakefield M Transcript length bias in RNA seq data confounds systems biology Biol Direct 2009 4 14 22 ExpressionPlot http expressionplot com 23 ExpressionPlot User s Guide http expressionplot com wiki 24 ExpressionPlot Google Group http groups google com group expressionplotl 25 CRAN Package Hmisc http cran r project org web packages Hmisc index html Friedman and Maniatis Genome Biology 2011 12 R69 Page 11 of 11 http genomebiology com 201 1 12 7 R69 26 BradStats R expressionplot Project Hosting on Google Code http code google com p expressionplot source browse trunk lib R BradStats R 27 European Nucleotide Archive ERP000619 http www ebi ac uk ena data view ERP000619 28 Affymetrix Sample Data Exon 1 0 ST Array Dataset http www affymetrix com support technical sample_data exon_array_data affx 29 Akira S Takeda K Toll like receptor signalling Nat Rev Immunol 2004 4 499 511 doi 10 1186 gb 201 1 12 7 r69 Cite this article as Friedman and Maniatis ExpressionPlot a web based framework for analysis of RNA Seq and microarray gene expression data Genome Biology 2011 12 R69 Submit your next manuscri
19. gh quality software packages now exist to perform specific steps in the analysis pipeline 4 10 as well as web based systems such as Galaxy 11 and GenePattern 12 that enable the management of data flow through these tools We present ExpressionPlot an open source solu tion consisting of a back end pipeline which performs alignment and statistical analyses and a web based front end which allows users to explore and further compare the completed analyses Compared to Galaxy and Gene Pattern ExpressionPlot s web based front end is novel in the ease with which one can browse and manipulate gene expression results gene isoform lists are one click filterable sortable and hyperlinked to the underlying genomic regions in the table_browser tool Furthermore even with differing platforms such as microarray versus RNA Seq or organisms such as mouse versus human the front end can automatically compare changes in gene expression across different experiments using the 4 way and heatmap tools Correspondence brad aaron riedman gmail com Department of Molecular and Cell Biology Harvard University 7 Divinity Ave Cambridge MA 02138 USA Full list of author information is available at the end of the article C BioMed Central ExpressionPlot can be tested as a virtual machine running under VirtualBox or installed directly into an existing web server Input to ExpressionPlot can be raw sequence data FASTQ files or Affyme
20. ide comparison of changes from different experiments data sets Specific tasks a examining reads probe intensities from a particular genomic region b examin ing levels changes of a particular gene splicing event or set of genes splicing events ExpressionPlot provides simple mechanisms to per form all of these steps Back end pre processing tasks RNA Seq Alignment ExpressionPlot uses bowtie 9 to align reads to the gen ome and then a database of splice junctions The splice junction databases that come with ExpressionPlot were generated by combining the known half junctions from each gene in every possible forward splicing combina tion exon n splices to exon m where m gt n Pre com puted junction databases can be downloaded and installed with the EP manage pl script human mouse and rat as of press time or can easily be generated using the make_junctions_database pl script that comes with ExpressionPlot ExpressionPlot s alignment strategy is to find and use only unique best alignments either to the genome or to the splice junction database Figure S1 in Additional file 1 For paired end data an additional step is taken to try to align the single ends individually Figure S2 in Additional file 1 Counting reads for genes and RNA processing events Aligned reads are then mapped to gene models and alternative splicing events Users can supply their own models and events or download and install pre com puted annotations
21. lues are still valid in these regimes the inclusion skip ratio statistic is less precise c Partial screen shot of table browser showing brain enriched cassette exons in the same data set The context menu was triggered by the mouse clicking on the row for CLTA clathrin light chain A and offers the user links to open the seqview genome browser tool in a window covering either the entire gene or just the alternative exon In either case the exon will be automatically highlighted Figure 5 Friedman and Maniatis Genome Biology 2011 12 R69 http genomebiology com 201 1 12 7 R69 After the plot is generated action buttons are pre sented to the user to access the significantly changed genes or RNA processing events in the table browser This screen presents the user with a dynamic table whose rows correspond to changed genes events Figure 3c The columns of the table contain identifiers for the gene or event like gene name chromsome strand and position as well as all the associated statistics such as read numbers RPKM values reads per kilobase gene model per million total reads and P values The table can be sorted by clicking on the header of the desired field or filtered using a text string or a numeric filter Action buttons allow for the export of the table into other software such as R or OpenOffice or Excel for automatic conversion of the genes into other IDs such as Ensembl or Entrez and for the automatic generation
22. m the ALS Therapy Alliance Author details Department of Molecular and Cell Biology Harvard University 7 Divinity Ave Cambridge MA 02138 USA The Koch Institute for Integrative Cancer Research Massachusetts Institute of Technology Cambridge MA 02139 USA Department of Bioinformatics and Computational Biology Genentech Inc 1 DNA Way South San Francisco CA 94080 USA Department of Biochemistry and Molecular Biophysics Columbia University College of Physicians and Surgeons 701 West 168th St New York NY 10032 USA Authors contributions BF conceived of and wrote the software and the manuscript TM helped in its design and coordination and in drafting the manuscript Both authors read and approved the final manuscript Competing interests The authors declare that they have no competing interests Received 24 December 2010 Accepted 28 July 2011 Published 28 July 2011 Page 10 of 11 References 1 Wang ET Sandberg R Luo S Khrebtukova Zhang L Mayr C Kingsmore SF Schroth GP Burge CB Alternative isoform regulation in human tissue transcriptomes Nature 2008 456 470 476 2 Nagalakshmi U Waern K Snyder M RNA Seq a method for comprehensive transcriptome analysis Curr Protoc Mol Biol 2010 Chapter 4 Unit 4 11 1 13 3 Mortazavi A Williams BA McCue K Schaeffer L Wold B Mapping and quantifying mammalian transcriptomes by RNA Seq Nat Methods 2008 5 621 628 4 Trapnell C Pachter L Salz
23. nd and so on as well as all the associated statistics It has the same fields that would be shown in the 2way browser but they are then repeated for both experiments This includes the annotation fields since sometimes they are from different organisms As with the 2way browser there are action buttons to download convert IDs and generate background sets Finally clicking on a row of the table opens a context menu with links that will automatically open the genome browser to the right part of the genome for the two experiments In the case of RNA processing events the correct genomic region will be automatically highlighted within the browser so the user can quickly find for example a differentially spliced cassette exon Page 7 of 11 The heatmap tool Figure S8 in Additional file 1 allows the user to compare larger numbers of change profiles Here all the different comparisons from one project are laid out along the x axis and all the com parisons from a second possibly different project are laid out along the y axis The color of each square of the heatmap indicates the similarity of the two com parisons The user can choose from a variety of statis tics to quantify similarity This tool is a useful way to look for relationships within larger numbers of experiments Web based front end specific tasks Examining reads from a particular genomic region The seqview tool is ExpressionPlot s genome browser Figure 5 With i
24. nd S9 in Addi tional file 1 are available from the European Nucleotide Archive under accession number ERP000619 available at 27 Archival copy of software For archival purposes version 1 3 of the software is included as Additional file 4 but it is recommended to use the latest version available through the website Additional material Additional file 1 Supplementary figures methods references and description of other additional files Additional file 2 Data for Figure S7 in Additional file 1 Additional file 3 Data for Figure 2d Additional file 4 Archival copy of software Abbreviations ECDF empirical cumulative distribution function PAC poly adenylation cleavage site RPKM reads per kilobase gene model per million total reads TSS transcription start site Acknowledgements We would like to thank Y Katz SL Ng J Gertz and M Muratet for critical reading of the manuscript S O Keeffe M Muratet and D Quest for software testing and technical suggestions CB Burge for hosting our prototype server D Housman for scientific advice and laboratory space during the development of this software IK Friedman and B Lewis for administrative support HP Phatnani C Lobsiger J Cahoy J Zamanian and other members of the Barres Lab Stanford Myers Lab HudsonAlpha Institute Ravits Lab Benaroya Institute and Maniatis Lab Harvard Columbia for providing data and or user feedback This work was supported by a grant fro
25. nome Biology 2011 12 R69 http genomebiology com 201 1 12 7 R69 The pairplot tool is a genome browser specifically designed to visualize the relationship between the aligned positions of paired ends Only one sample can be visualized at a time The gene annotation of the requested region is shown as well as the pileup track from the seqview tool showing total numbers of reads Above this a scattergram shows a point for each paired end read aligning to the genomic region The x axis gives the position of the plus strand end and the y axis gives the position of the minus strand end The colors and sizes of the points indicate the number of reads aligning to each pair of coordinates Under conditions of constitutive splicing the scattergram should form a ser ies of segments above each exon and parallel to the diagonal with the distance to the diagonal dictated by the paired end insert and intron size Alternatively spliced regions however will show multiple parallel seg ments corresponding to the different isoforms The rela tive strength of the segments corresponds to the abundances of the two isoforms Figure S9 in Additional file 1 Examining levels or changes of particular genes or events The genelev tool generates barplots of gene levels RPKM with error bars Figure 6a The ecdf tool allows the user to visualize the levels or fold changes of a set of genes by plotting the cumulative distribution of those genes levels in the s
26. orrespond to genes Fold change of expression in heart is plotted versus all other samples in corresponding data set Genes enriched in heart are plotted further to the right exon array and or up RNA Seq and those higher in other samples are further to the left and or down Genes significantly different only on one platform are colored red exon array or green RNA Seq and those different on both platforms are colored blue P value cutoffs are 0 01 for exon array and 10 for RNA Seq and fold change cutoffs are 2 for both platforms Colored numbers show number of genes in each category b Similar plot comparing the same x axis human heart enriched gene expression by exon array to mouse heart enriched gene expression also by exon array y axis Friedman and Maniatis Genome Biology 2011 12 R69 http genomebiology com 201 1 12 7 R69 decreased in the x axis experiment but increased in the y axis experiment Points are colored according to whether the gene event is significantly changed in one or both experiments with blue representing those chan ged in both experiments As with the 2way tool after the plot is generated ExpressionPlot offers the user action buttons to select a group of genes events to further examine in the 4way table browser For example clicking Up Up would show a table of genes events increased in both experi ments This table shows the annotation of the gene event identifier chromosome position stra
27. pt to BioMed Central and take full advantage of e Convenient online submission e Thorough peer review e No space constraints or color figure charges e Immediate publication on acceptance e Inclusion in PubMed CAS Scopus and Google Scholar e Research which is freely available for redistribution Submit your manuscript at C www biomedcentral com submit BioMed Central
28. r in the cerebellum row min C q2 0 lt q2 lt 1 Solving the problem by convex optimization methods would be feasible but slow due to the cost of re calcu lating C q2 Instead we use the binconf function from R s Hmisc library 25 to calculate a 95 confidence interval for q2 for every gene based on the observed number of reads This interval corresponds to the range of q2 for which that gene is not significantly changed Then the range 0 to 1 is split into windows of width 0 0001 and the number of genes whose confidence interval overlaps each of these windows is counted The uncertainty introduced by using windows as point esti mates is mitigated by their small radius a difference of 0 0001 0 01 in the sample size estimate will have a minute effect on resultant gene levels The value of q3 for the window overlapped by the confidence intervals of the most genes or the mean of the q for the several windows if there is a tie for the most intervals is then taken as the optimum Empirical tests show that this method is extremely robust to the choice of P value cut off data not shown This is implemented in a very Friedman and Maniatis Genome Biology 2011 12 R69 http genomebiology com 201 1 12 7 R69 short R function called minimize significant changes in BradStats R 26 European Nucleotide Archive accession numbers The previously unpublished and de identified data sets used to create Figure 2d and Figures S7 a
29. scle SkelMuscle Fraction of Ganonical Paired End Reads 0 20 Paired End Distance Figure 2 Screen shots of ExpressionPlot quality control tools a read_types tool showing all read types Numbers of non aligning Nonmatch mulitply aligning Mult unique genome aligning Genomic and unique junction aligning Junction reads are shown for each lane from a mouse tissue transcriptome dataset 3 Numbers 1 2 indicate different libraries letters A B C indicate different lanes of the same library b read_types tool showing matching read types normalized to 100 c Pairwise correlation heatmap of gene expression profiles generated from each lane d pairdist tool shows ECDF of paired end distances of canonical reads same chromosome different strand minus strand read downstream of plus strand read Distance is defined as the genomic distance in nucleotides between the aligned positions of the last sequenced bases of the two reads can be negative if the alignments overlap The samples have been de identified data in Additional file 3 Numbers in parentheses indicate median paired end distance for each sample add 36 for both sequences and 50 for both Illumina adaptors 172 to get complete library size Friedman and Maniatis Genome Biology 2011 12 R69 http genomebiology com 201 1 12 7 R69 plus end strand and 4 the two ends align to the same chromosome different strands minus end downstream of the plus
30. t the user can select the project of interest then query either by a gene name or genomic region One of several annotations can be chosen and then a plot is generated showing either the pileup of reads in that region with strands separated or merged as requested by the user or of the hybridization intensi ties of microarray probes in that region Zooming and scrolling is implemented and users can also highlight specific genomic coordinates Barplots are automatically generated showing levels of genes within the requested regions 200 brain a No Reads seer breast No Reads an 4 liver aa No Reads o 200 heart E No Reads an f skel_muscle a No Reads a i protein_coding CLTA protein_coding GLTA protein_coding CLTA H gt pratein_coding CLTA 36 165 000 36 190 000 36 195 000 36 200 000 ghra brain enriched exon F Figure 5 Screen shot of ExpressionPlot s genome browser seqview The region of the CLTA gene which contains a brain enriched exon pink is shown Known transcripts of CLTA are seen along the bottom arrowheads indicate plus strand The accumulation of RNA Seq reads from five human tissues is shown on the top The heights of black bars indicate numbers of reads overlapping each genomic position whereas the heights of blue brackets indicate numbers of reads overlapping splice junctions Data from RNA Seq human tissue panel 1 Friedman and Maniatis Ge
31. trix array data CEL files completed alignments BAM files or tables of gene expression values and changes generated by other back ends Once data are pre processed the web based front end allows users to easily browse measures of quality control plot changes in gene expression and RNA processing browse hyperlinked tables of changed genes and splicing events generate read plots from a genomic view compare different datasets including from different organisms or between microarray and RNA Seq generate empirical cumulative distribution functions ECDFs to look at levels or changes in a cohort of genes and look up levels of specific genes The ExpressionPlot back end can also generate BAM and BigWig files upon request and for downstream ana lysis the web based front end can output spreadsheets with gene and exon statistics ExpressionPlot includes a web controllable user account and access control system by which pre published data can be shared with other users or when appropriate made public Finally ExpressionPlot does not require a cluster it can run on any machine with sufficient memory to hold the bowtie indexes usually at least 3 or 4 GB and hard drive space to hold the sequencing data and processed files roughly 1 to 2 GB per lane In short ExpressionPlot is a unified solution for gene expression analysis of RNA Seq and microarray data 2011 Friedman and Maniatis licensee BioMed Central Ltd This is an open ac
32. using EP manage pl currently avail able for human mouse and rat The pre computed Page 2 of 11 gene models are built from all exons of any transcript based on UCSC known genes 13 or Ensembl 14 A read is counted towards any gene that contains the aligned positions possibly split by a junction on either strand within its exons Scripts and detailed instructions to generate annotations for other genomes are included Pre computed candidate skipped exon events are cre ated from all known exons regardless of whether or not they are known to be skipped For skipped exons skip ping reads are considered as splice junction spanning reads that both skip the exon and are additionally anchored in known splice sites of the host genes Figure S3 in Additional file 1 For intron retention the number of reads aligning to the intron is compared to the number aligning to locally constitutive flanking exons Figure S4 in Additional file 1 Locally constitutive means that based on the under lying annotation all transcripts flanking that intron con tain those exons Figure S5 in Additional file 1 As with skipped exons the pre computed sets contain candidate events for all known introns Finally alternative terminal exon events are created for genes with multiple transcript start sites TSSs or multiple poly adenylation cleavage sites PACS These events compare reads supporting a candidate terminal exon with more distal 5 of TSS
Download Pdf Manuals
Related Search