Gao et al.NIH-PA Author ManuscriptNIH-PA Author ManuscriptNIH-PA Author ManuscriptPage 11
interactions from the Human Reference Protein Database (HPRD) (17), Reactome (18),National Cancer Insititue (NCI)–Nature (19), and the Memorial Sloan-Kettering CancerCenter (MSKCC) Cancer Cell Map (http://cancer.cellmap.org), as derived from the opensource Pathway Commons Project (20). By default, the network that is automatically
generated contains all neighbors of all query genes. If more than 50 neighbor genes exist inthe network, they are ranked by genomic alteration frequency within the selected cancerstudy, and only the 50 neighbors with the highest alteration frequency in addition to thequery genes are shown. This provides an effective means of managing network complexityand automatically highlights the genes most relevant to the cancer type in question. The full,nonpruned network can be downloaded in the SIF (simple integration file) and GraphMLformats for visualization and analysis in Cytoscape (21). By default, the portal automaticallycolor codes edges by interaction type and overlays multidimensional genomic data onto eachnode, highlighting the frequency of alteration by mutation, CNA, and mRNA up- or down-regulation. The data that are shown depend on the settings used in the query and the datathat are available for the selected genomic profiles. Various options for filtering the networkare available, and the network can be searched by gene symbol. Various options for alteringthe display of the network and the layout of the network are available. Legends explainingthe network symbols are provided. Details about the alterations found in the genes and theinteractions between the genes are viewed by clicking on the node or the edge, respectively.Interaction types are derived from the BioPAX to SIF inference rules (20). For example, “InSame Component” indicates that Genes A and B are involved in the same biological
component, such as a complex. “State Change” indicates that Gene A causes a state change,such as a phosphorylation change, within Gene B. “Other” is used to indicate all other typesof inter actions, including protein-protein interactions derived from HPRD. “Targeted byDrug” indicates a drug-target interaction.
The portal contains gene-centric drug-target information from the following resources:
DrugBank (22), KEGG Drug (23), NCI Cancer Drugs (http://www.cancer.gov/cancertopics/druginfo/alphalist), and Rask-Andersen et al. (24). Drugs are hidden from the network
display by default but can be added to the network by using the Genes & Drugs menu. Usershave the option of displaying U.S. Food and Drug Administration (FDA)–approved drugs,cancer drugs defined by NCI Cancer Drugs, or all drugs targeting the query genesNew networks can be generated by selecting genes in the current network and thensubmitting those genes as a new query.
For example, to identify genomic alterations in epidermal growth factor receptor (EGFR)signaling networks in serous ovarian cancer, we used EGFR and ERBB2 as the query genesand explored the resulting network (Fig. 8). Using the color-coding as a guide, connectedgenes with alterations in this cancer are obvious. For the EFGR and ERBB2 network MYC,a known downstream effector of ERBB2 (25), is colored more intensely red because it isamplified in 30% of the TCGA ovarian cancer samples (Fig. 8).
By adding the drug data, gefitinib and erlotinib, which are tyrosine kinase inhibitors thattarget the catalytic domain of EGFR, and cetuximab and trastuzumab, which are monoclonal
Sci Signal. Author manuscript; available in PMC 2014 September 10.
Gao et al.NIH-PA Author ManuscriptNIH-PA Author ManuscriptNIH-PA Author ManuscriptPage 12
antibodies that target the extracellular domain of EGFR and ERBB2, respectively, showwith edges connecting them to their targets (Fig. 8A) (26, 27).
1.Perform the query shown in Fig. 8.2.Select the Network tab.
3.Select “Show all Drugs” from the Genes & Drugs tab.
4.From the Layout button, select “Layout Properties” and set the maximum distanceto 100 to shorten the length of the edges.5.From the Layout button, select “Perform Layout.”
6.To automatically perform layout changes after filtering the network, select “Autolayout on changes.”
7.Set the “Filter Neighbors by Alteration” to 10.
8.Rearrange nodes by single clicking and repositioning nodes for better layout.9.
Double click the MYC node to view genomic profile details.
10.From the View menu, select “Highlight neighbors,” then select “Remove
highlights” to restore all nodes and edges.11.View and filter interaction types and sources in the Interactions tab.12.Double click the line connecting Flavopiridol to EGFR to view details.
13.Deselect “Merge Interactions” to show multiple edges of different interaction types
between nodes.14.From the View button, select “Always Show Profile Data” to visualize the
alteration frequencies of different genomic profiles around each gene. Deselect toremove.15.Use the options from the “Topology” button to hide or show only selected nodes or
remove disconnected nodes from the network.16.Select EGFR, ERBB2, and MYC from the Genes & Drugs tab and click the arrow
button to submit a new query.17.Use the browser back button to return to the previous result.
18.Download GraphML or SIF for further analysis in other tools such as Cytoscape.Results Tabs 8-10: IGV, Download, Bookmark—The Download tab provides allgenomic data and per-sample alteration events for download. Users can download tab-delimited text files with all data for the query genes or simply copy event information intoan external spreadsheet application for further analysis. The tab-delimited text files are
available in two formats: (i) a data matrix of genes (rows) versus samples (columns) and (ii)a transposed matrix of samples (rows) versus genes (columns).
Sci Signal. Author manuscript; available in PMC 2014 September 10.
Gao et al.Page 13
Users can also visualize copy number details by choosing to launch a Web start version ofthe IGV (28). IGV will open the segmented copy-number data of the current cancer studyand display the copy-number status of all query genes.
The Bookmark tab allows users to save or bookmark a specific query (the entire query canbe stored in a URL) or share their results with collaborators by generating a short URL(using bit.ly).
1.2.
Perform any query.
From the IGV tab, click the “Launch” button to load the data and start the viewer.
Note: The segmented copy-number data for all samples are visualized in IGV,regardless of which cases are selected for querying in the cBioPortal.
3.
From the Download tab, to obtain the data in tab-delimited format, click the
hyperlinks to view the file desired or open the URL in a new tab or window. Then“select all” to copy into a spreadsheet or select “File,” then “Save Page as” to saveas a text file.
From the Download tab, to place the data into a spreadsheet or create a filemanually, copy and paste the data in each text box into the program of choice.From the Bookmark tab, right-click (on a PC) the link shown and paste into abrowser to create a personal bookmark or to store the link to the specified query.From the Bookmark tab, press the “Shorten URL” button to create a shorter URLfor the specified query using bit.ly.
Note: Clicking on the short link or the long version will reload the Bookmarktab page for the specified query.
Performing Cross-Cancer Queries
Cross-cancer queries allow users to assess alteration frequencies and mutation data forindividual genes or combinations of genes across multiple different cancer types. Cross-cancer queries of mRNA expression or protein abundance data are not yet available. Theportal will automatically limit the studies searched to match the query parameters so thatonly data with mutation information is included for a mutation-only query and only datawith CNA information is included for a CNA-only query. The results are presented as ahistogram: (i) one showing the frequency of the alterations in the cancers, which can bepresented in descending order; or (ii) one showing the absolute number of samples with andwithout alterations in each cancer study, which can be presented in order of decreasing
number of cases with alterations. If multiple genes are queried, then the histograms representthe combined alterations or alteration frequency in all of the selected genes. Detailsregarding the queried genes in the form of OncoPrints for each cancer study are alsoprovided. This enables the results for each selected gene to be visualized for each cancerstudy.
A cross-cancer query of TP53, which encodes the tumor suppressor and transcriptionalregulator p53, illustrates this feature of the cBioPortal (Fig. 9A).
NIH-PA Author ManuscriptNIH-PA Author ManuscriptNIH-PA Author Manuscript4.5.6.
Sci Signal. Author manuscript; available in PMC 2014 September 10.
Gao et al.Page 14
1.
General and Specific: Select “All Cancer Studies” from the main query page(Home).
General: Select data types.Specific: Select “Only Mutation.”
Note: This will automatically limit the query to only those cancer studies withmutation data.
NIH-PA Author ManuscriptNIH-PA Author ManuscriptNIH-PA Author Manuscript2.
3.General: Enter genes of interest.Specific: Enter TP53.
4.5.6.
Press “Submit.”
Press the “Sort” link to organize the data from cancers with the most to those withthe least frequently occurring mutations in the query gene (Fig. 9B).
To view the data as the absolute number of altered and unaltered samples, select“Show number of altered cases (studies with mutation data)” from the drop-downmenu.
Mouse over any bar in the histogram to view a summary of the results.
Click the arrowhead beside any of the listed cancer studies to view the OncoPrintsfor the selected genes.
Click “View Cancer Study Details” to execute the query in the selected cancerstudy, which enables access to all of the results listed for a single study query.
7.8.9.
10.Use the browser back button to return to the cross-cancer query results.11.Click the “Export” link to download the data as an SVG file.
Viewing Cancer Study Summary Data
In addition to performing specific gene queries, the cBioPortal provides access to summaryinformation about each cancer study included in the portal. The data available includevarious clinical details about the patients (survival and age at diagnosis), details about thetumor (histology, stage, grade), and summaries of the genomic data (number of
nonsynonymous mutations and fraction of genome altered), details about the recurrentlymutated genes, and details about recurrent CNAs. The clinical data are presented bothgraphically and in table format (Fig. 10). The mutated gene and CNA data are presented intables. All tables have a search option. The search queries all content (case IDs, genesymbols, and clinical attributes) in the table containing the searched term or phrase.
1.2.3.4.
Select “Uterine Corpus Endometroid Carcinoma (TCGA, Provisional)” from thedrop-down menu in the main query page (HOME).Press the “View details” button.
Press the “more?” button to see additional graphical summaries.Mouse over the data in the graphical summaries for details.
Sci Signal. Author manuscript; available in PMC 2014 September 10.
Gao et al.Page 15
5.
Sort the data in the clinical data table by clicking the arrowheads next to eachcolumn. Use the scroll bars to move up and down or across the table.
Search for deceased patients by typing “Deceased” (without quotations) into thesearch box.
Note: Searching the table of patient data below the graphical summaries willnot update the graphical data for the selected patient.
NIH-PA Author ManuscriptNIH-PA Author ManuscriptNIH-PA Author Manuscript6.
7.8.9.
Restore the full list of cases by deleting the search text from the search box.Click the tab “Copy Number Alterations” to access a list of chromosomal regionsand genes with CNAs.
Click the tab “Mutated Genes” to access the list of recurrently mutated genes.
10.Click any of the listed genes to execute a new query for mutations of the selected
gene in the selected cancer study.11.Use the browser back button to return to the cancer study summary, which displays
the “Clinical Data” results.12.Click the “Serous” pie in the Histology pie chart to update other plots and the table
to reflect the results of only those cases that are of the serous type.13.Click the “Clear selection” button to restore all plots and table.
Viewing Genomic Alterations in a Single Tumor: Patient View
Because there are potentially hundreds or thousands of genomic alterations in any singletumor sample, it is crucially important to select, for inspection and analysis, alteration eventsthat most likely contribute to oncogenesis or affect the response to therapy. Therefore, inaddition to gene-by-gene alteration maps across many samples and across diverse tumortypes and the cancer study summary data, users can also view genomic alterations inindividual tumor samples in an interactive patient view page. Links to these pages are
available from the OncoPrint (through the mouse-over details for each genomic event), theMutations tab, and the cancer study summary page.
The patient view summarizes and visualizes all relevant data about a tumor, includingclinical characteristics, summaries of the extent of mutations and copy-number alterations,as well as details about mutated, amplified, and deleted genes (Fig. 11). The results aredisplayed in tabbed displays. Genomic alterations in the summary tab are filtered by thefollowing criteria: recurrence of mutations or CNAs across the tumor cohort (from MutSigand GISTIC), mutation occurrence in COSMIC (14), and cancer gene annotation [fromresources, such as the Sanger Cancer Gene Census (29)]. The patient view also providesinformation about drugs that target the altered genes and lists relevant clinical trials fromhttp://www.cancer.gov/.
1.2.3.
Click the “DATA SETS” button at the top of the navigation pane.Click “Uterine Corpus Endometrioid Carcinoma (TCGA, Provisional).”Enter “TCGA-FI-A2D2” in the search box above the table.
Sci Signal. Author manuscript; available in PMC 2014 September 10.