Gao et al.NIH-PA Author ManuscriptNIH-PA Author ManuscriptNIH-PA Author ManuscriptPage 6
cases without an alteration by selecting “Remove Unaltered Cases”; and (iii) select“Remove Whitespace” to eliminate the gaps between samples.
4.To restore the original case order (alphabetically by case ID or as defined by theuser in the original query), select “Restore Case Order” in the “Customize” options.5.To export the OncoPrint, choose to download the OncoPrint as an XML file inscalable vector graphic (SVG) format by pressing the SVG button.
6.To obtain additional information, mouse over the indicated alteration on the gene.7.
To modify or start a query, choose “Modify Query” above the tabs for the results.
Results Tab 2: Mutual Exclusivity—Biological processes or pathways in cancer areoften deregulated through different genes or by multiple different mechanisms. The conceptof mutual exclusivity can be exploited to identify previously unknown mechanisms thatcontribute to oncogenesis and cancer progression (12). In mutual exclusivity, events ingenes associated with a specific cancer tend to be mutually exclusive across a set of tumors—that is, each tumor is likely to have only one of the genetic events. The opposite situation(co-occurrence) is when genetic alterations occur in multiple genes in the same cancersample. The portal computes a set of simple statistics to identify patterns of mutual
exclusivity or co-occurrence. For each pair of query genes (G1 and G2), the portal calculatesan odds ratio (OR) (Eq. 1) that indicates the likelihood that the events in the two genes aremutually exclusive or co-occurrent across the selected cases:
(1)
Where A = number of cases altered in both genes; B = number of cases altered in G1 but notG2; C = number of cases altered in G2 but not G1; and D = number of cases altered inneither genes.
It then assigns each pair to one of five categories that are indicative of a tendency towardmutual exclusivity, of a tendency toward co-occurrence, or of no association. A legend isprovided with the analysis. To determine whether the identified relationship is significant foreach gene pair, the portal performs a Fisher's exact test.
Using the same query used for describing OncoPrints, the mutual exclusivity analysis showsthat events in the three selected genes tended to occur in a mutually exclusive way, but thepattern was only statistically significant for CDKN2A and CDK4, and for CDKN2A andRB1, but not for CDK4 and RB1, which may be due to the small sample size (Fig. 3). Thisfits with what is known about RB signaling in GBM, which can be deactivated by
inactivation of RB1 itself (through mutation or deletion), by activation of CDK4 (a CDKthat inhibits RB1 activity) through amplification, or by inactivation of the CDK inhibitorp16, which is encoded by CDKN2A, through deletion or mutation. Thus, a single alterationin one of these genes is sufficient to deactivate the pathway, and this is what the mutualexclusivity analysis showed.
Sci Signal. Author manuscript; available in PMC 2014 September 10.
Gao et al.NIH-PA Author ManuscriptNIH-PA Author ManuscriptNIH-PA Author ManuscriptPage 7
1.
Perform the query as specified in Fig. 3. Once the “submit” button is pressed, theOncoPrint result is displayed automatically.2.
Select the Mutual Exclusivity tab.
Note: This tab will only show if more than one gene is selected in the query.
Results Tab 3: Correlation Plots—The cBioPortal offers several different ways ofvisualizing discrete genetic events (CNAs or mutations) and continuous events, such as dataregarding mRNA or protein abundance, or DNA methylation.
For each gene specified in the query, the portal can generate various plots, depending on thedata available. The mRNA versus copy-number option displays a box-and-whisker plot toshow mRNA expression from user-selected data sources of a gene plotted in relation to itscopy-number status in each sample. Copy-number status can be homozygously deleted,heterozygously deleted, diploid, gained (meaning an amplification event with relatively fewcopies), or amplified (meaning an amplification event with many copies). The mRNA-versus-DNA methylation option displays a scatter plot of mRNA expression compared withDNA methylation data of a gene across all selected samples. A methylation beta-value is anestimate for the methylation level of a CpG locus using the ratio of intensities between
methylated and unmethylated alleles. The RPPA protein level versus mRNA option displaysa scatter plot of protein abundance compared with mRNA abundance for a gene across allselected samples.
Genes and data types are selected by using drop-down menus, and only those options forwhich data are available are provided in the menus. All plots can be exported as PDFdocuments for use in publications.
The example query to illustrate this type of analysis is a query of ERBB2 (a known proto-oncogene encoding an epidermal growth factor receptor) in colon and rectum
adenocarcinoma. ERBB2 is amplified in a subset of colorectal cancer samples (8). ThecBioPortal results show that ERBB2 mRNA is increased in the samples in which ERBB2 isamplified (Fig. 4A) and that the tumors with the highest amount of ERBB2 mRNA had thehighest amount of ERBB2 protein (Fig. 4B).
1.Perform the query shown in Fig. 4. Once the “submit” button is pressed, theOncoPrint result is displayed automatically.2.Select the Plots tab.
3.Select “mRNA expression (microarray)” from the first Data Types menu.4.Select “Putative copy-number alternations from GISTIC” from the second DataTypes menu.
5.Select “mRNA v. Copy Number” from the Plot Type menu.6.Press the arrow button to generate the graph shown in Fig. 4A.7.To export as a PDF, click the PDF link at the top near the graph title.8.
Select “RPPA protein level v. mRNA” from the Plot Type menu.
Sci Signal. Author manuscript; available in PMC 2014 September 10.
Gao et al.NIH-PA Author ManuscriptNIH-PA Author ManuscriptNIH-PA Author ManuscriptPage 8
9.Press the arrow button to generate the graph shown in Fig. 4B.
Note: If a combination that cannot be plotted is selected, an error message isdisplayed.
Results Tab 4: Mutations—The Mutations tab provides details as both a graphicalsummary and a customizable table about all nonsynonymous mutations identified in eachquery gene. The graphical summary shows the position and frequency of all mutations in thecontext of Pfam protein domains (13) encoded by the canonical gene isoform. All DNAmutations are standardized to the canonical RefSeq isoform (using Oncotator, http://www.broadinstitute.org/oncotator/). When a DNA mutation only affects noncanonicalisoforms, the mutations are not included in the graphical summary. Future versions of theportal will provide this information in a separate table.
Below the graphical summary is a table of all nonsynonymous mutations. This table, whichcan be sorted and filtered, provides the following information if the data are available: caseID for each sample (hyperlinked to the patient view page of the specific sample containingthe mutation); amino acid change; type of mutations (missense, nonsense, splice site,
frameshift insertion or deletion, in-frame insertion or deletion, nonstop, nonstart); number ofmutations at this position in COSMIC (Catalogue Of Somatic Mutations In Cancer) (14);predicted functional impact of missense mutations [with hyperlinks to Mutation Assessor(15) for the specified mutation and a multiple sequence alignment]; link to a 3D structurewith the mutation highlighted (with hyperlinks to Mutation Assessor); mutation status(somatic or germline–germline mutations are currently only provided for BRCA1 and
BRCA2 in some studies); validation status (valid or unknown); the sequencing center wherethe sample was sequenced and the mutation identified; variant allele frequency in the tumor;variant allele frequency in the matched normal sample; exact genomic position
(chromosome, start, end, reference allele, variant allele); variant and reference allele counts(the number of variant and reference alleles found in the sequencing results of tumor andnormal samples); and information about the affected isoform. The last three are not shownby default but may be displayed. Users can perform a search for any text in the table withthe search option.
The example query to illustrate this type of analysis is a query of ERBB2 in colon andrectum adenocarcinoma using only sequenced tumors (Fig. 5). The graphical summary ofthe mutations associated with this query showed that there are 10 ERBB2 nonsynonymousmutations in colorectal cancer samples, and four of them are V842I in the kinase domain(Fig. 5), suggesting that this is a hotspot for protein activation. From the table, the kinasedomain mutations at amino acids 755, 777, and 842 have been observed in several othercancer studies before (6, 8, and 2 COSMIC entries, respectively) (Fig. 5B).
1.Perform the query shown in Fig. 5.2.Select the Mutations tab.
3.
Mouse over the colored regions representing protein domains to view details aboutthe domain and its starting and ending residues in the protein sequence.
Sci Signal. Author manuscript; available in PMC 2014 September 10.
Gao et al.NIH-PA Author ManuscriptNIH-PA Author ManuscriptNIH-PA Author ManuscriptPage 9
4.
Mouse over the circles above the protein sequence diagram to see the specificmutation. The height of the line connecting the circle to the protein is indicative ofthe frequency of the mutation. The most frequent mutation is labeled with its aminoacid change.
5.
Customize the data displayed in the table using the “Show/hide columns” menu.Select those to display. Deselect those to hide.
Note: The following columns are hidden by default: Exact genomic position(chromosome, start, end, reference, variant allele); variant and reference alleleread counts in tumor and normal samples; and information about the affectedisoform.
6.Use the up and down arrowheads to sort the data according to the column values.7.Follow the hyperlinked Case ID to get details about the tumor sample containingthe mutation.
8.Use the browser back button to return to the Mutations tab.
9.
Mouse over the values in the COSMIC column to get details about the frequencyand specific mutations at that residue.
10.Mouse over the values in the FIS column to follow hyperlinks to the Mutation
Assessor or a Multiple Sequence Alignment.11.Click the 3D link to view 3D protein structures with the mutated amino acid
highlighted and return to the Mutations tab by using the browser back button.12.Enter “V842I” (without quotations) in the search box to display V842I mutations
only.
Note: The search options in tables in the cBioPortal support free text search onthe table content.
13.Delete the search text to return to the complete results.
Results Tab 5: Protein Changes—Protein and phosphoprotein data are available fromthe Protein Changes tab. Currently, large-scale proteomics data from the RPPA (16)platform are available in the portal for 12 TCGA cancer studies (table S1). As already
described, scatter plots of protein abundance versus mRNA expression for query genes canbe generated if both data types are available (Fig. 4B, Plots tab).
For each query, the portal also performs differential analysis for all available RPPA proteindata and identifies protein and phosphoprotein events that correlate with genomic alterationsin the query genes. It is not necessary to select “RRPA proten/phosphoprotein level” fromthe query screen. If the data are available, then this analysis can be performed. For eachavailable protein or phosphoprotein, cBioPortal performs a two-sided, two-sample Student'st test to identify differences in protein abundance between tumor samples that have at leastone event (alteration) in one of the query genes, and those that do not. The results aredisplayed as a list of proteins or phosphoproteins, ranked by their difference in abundancebetween altered and unaltered samples. The table includes the following information: the
Sci Signal. Author manuscript; available in PMC 2014 September 10.
Gao et al.NIH-PA Author ManuscriptNIH-PA Author ManuscriptNIH-PA Author ManuscriptPage 10
target protein recognized by the antibody; the residue phosphorylated or modified (for
example, by cleavage); the average protein abundance z scores in the tumors with alterationsand those without (unaltered); the P value; and an option to plot the results, which are shownby default. The RPPA ID and the absolute difference between the unaltered and alteredsamples' average z scores are optional columns. For each protein or phosphoprotein, the zscores of the RRPA data between the unaltered and altered samples can be displayed as abox plot.
A query of glioblastoma cancers for mutations and CNAs associated with the tumorsuppressor and lipid phosphatase encoded by PTEN illustrate this analysis (Fig. 6). Forexample, PTEN loss (mutation or copy number deletion) in glioblastoma cancer is tightlycorrelated with increased phosphorylation of AKT (pT308 and pS473) (Fig. 6).
1.Perform the query shown in Fig. 6.2.Select the Protein Changes tab.
3.Use the drop-down menu for “Antibody Type” to specify data collected usingantibodies that detect the total protein or the phosphoprotein.
4.Customize the data displayed in the table using the “Show/hide columns” menu.Select those to display. Deselect those to hide.
5.
Press the + symbol in the Plot column to display the boxplot comparing the z scoresfor abundance between the samples with alterations and those without alterations inthe queried gene (or genes).
6.Enter “ERBB” (without quotations) in the search box to display ERBB2 andERBB3 phosphoprotein changes.
7.
Delete search text to return to the complete results.
Results Tab 6: Survival—If survival data are available, overall survival and disease-freesurvival differences are computed between tumor samples that have at least one alteration inone of the query genes and tumor samples that do not. The results are displayed as Kaplan-Meier plots with P values from a logrank test.
A query for BRCA1 and BRCA2 mutations in ovarian cancer is used to illustrate theseresults. The analysis showed a significantly better overall and disease-free survival ofpatients with either a BRCA1 or BRCA2 mutation (Fig. 7).
1.Perform the query shown in Fig. 7.2.Select the Survival tab.
3.View the results for overall survival analysis and disease-free survival analysis.4.
Click the PDF link at the top near the title of each graph to download a PDFversion of the plot.
Results Tab 7: Network—The Network tab provides interactive analysis and
visualization of networks that are altered in cancer. The network consists of pathways and
Sci Signal. Author manuscript; available in PMC 2014 September 10.