In [ ]:
library(knitr)
opts_chunk$set(out.extra='style="display:block; margin: auto"', fig.align="center", tidy=TRUE)
This package provides a basic set of R functions for querying the Cancer Genomic Data Server (CGDS) hosted by the Computational Biology Center (cBio) at the Memorial Sloan-Kettering Cancer Center (MSKCC). This service is a part of the cBio Cancer Genomics Portal, http://www.cbioportal.org/.
In summary, the library can issue the following types of queries:
getCancerStudies(): What cancer studies are hosted on the server? For example, TCGA glioblastoma or TCGA ovarian cancer.
getGeneticProfiles(): What genetic profile types are available for cancer study X? For example, mRNA expression or copy number alterations.
getCaseLists(): what case sets are available for cancer study X? For example, all samples or only samples corresponding to a given cancer subtype.
getProfileData(): Retrieve slices of genomic data. For example, a client can retrieve all mutation data for PTEN and EGFR in TCGA glioblastoma.
getClinicalData(): Retrieve clinical data (e.g. patient survival time and age) for a given cancer study and list of cases.
Each of these functions will be briefly described in the following sections. The last part of this document includes some concrete examples of how to access and plot the data.
The purpose of this document is to give the reader a quick overview of the cgdsr package. Please refer to the corresponding R manual pages for a more detailed explanation of arguments and output for each function.
Initially, we will establish a connection to the public CGDS server hosted by Memorial Sloan-Kettering Cancer Center. The function for creating a CGDS connection object requires the URL of the CGDS server service, in this case http://www.cbioportal.org/, as an argument.
In [ ]:
library(cgdsr)
# Create CGDS object
mycgds <- CGDS("http://www.cbioportal.org/")
The variable mycgds is now a CGDS connection object pointing at the URL for the public CGDS server. This connection object must be included as an argument to all subsequent interface calls. Optionally, we can now perform a set of simple tests of the data returned from the CGDS connection object using the test function:
In [ ]:
# Test the CGDS endpoint URL using a few simple API tests
test(mycgds)
In [ ]:
# Get list of cancer studies at server
# NOTE: Only show the first few with head()
head(getCancerStudies(mycgds)[, c(1,2)])
Here we are only showing the first two columns, the cancer study ID and short name, of the result data frame. There is also a third column, a longer description of the cancer study. The cancer study ID must be used in subsequent interface calls to retrieve case lists and genetic data profiles (see below).
This function queries the CGDS API and returns the available genetic profiles, e.g. mutation or copy number profiles, stored about a specific cancer study. Below we list the current genetic profiles for the TCGA glioblastoma cancer study:
In [ ]:
head(getGeneticProfiles(mycgds,'gbm_tcga')[,c(1:2)])
Here we are only listing the first two columns, genetic profile ID and short name, of the resulting data frame. Please refer to the R manual pages for a more extended specification of the arguments and output.
This function queries the CGDS API and returns available case lists for a specific cancer study. For example, within a particular study, only some cases may have sequence data, and another subset of cases may have been sequenced and treated with a specific therapeutic protocol. Multiple case lists may be associated with each cancer study, and this method enables you to retrieve meta-data regarding all of these case lists. Below we list the current case lists for the TCGA glioblastoma cancer study:
In [ ]:
getCaseLists(mycgds,'gbm_tcga')[,c(1:2)]
Here we are only listing the first two columns, case list ID and short name, of the resulting data frame. Please refer to the R manual pages for a more extended specification of the arguments and output.
The function queries the CGDS API and returns data based on gene(s), genetic profile(s), and a case list. The function only allows specifying a list of genes and a single genetic profile, or oppositely a single gene and a list of genetic profiles. Importantly, the format of the output data frame depends on if a single or a list of genes was specified in the arguments. Below we are retrieving mRNA expression and copy number alteration genetic profiles for the NF1 gene in all samples of the TCGA glioblastoma cancer study:
In [ ]:
getProfileData(mycgds, "NF1", c("gbm_tcga_gistic","gbm_tcga_mrna"), "gbm_tcga_all")[c(1:5),]
We are here only showing the first five rows of the data frame. In the next example, we are retrieving mRNA expression data for the MDM2 and MDM4 genes:
In [ ]:
getProfileData(mycgds, c("MDM2","MDM4"), "gbm_tcga_mrna", "gbm_tcga_all")[c(1:5),]
We are again only showing the first five rows of the data frame.
The function queries the CGDS API and returns available clinical data (e.g. patient survival time and age) for a given case list. Results are returned in a data frame with a row for each case and a column for each clinical attribute. The available clinical attributes are:
Below we retrieve clinical data for the TCGA ovarian cancer dataset (only first five cases/rows are shown):
In [ ]:
getClinicalData(mycgds, "ova_all")[c(1:5),]
As a simple example, we will generate a plot of the association between copy number alteration (CNA) status and mRNA expression change for the NF1 tumor suprpressor gene in glioblastoma. This plot is very similar to Figure 2b in the TCGA research network paper on glioblastoma (McLendon et al. 2008). The mRNA expression of NF1 has been median adjusted on the gene level (by globally subtracting the median expression level of NF1 across all samples).
In [ ]:
df <- getProfileData(mycgds, "NF1", c("gbm_tcga_gistic","gbm_tcga_mrna"), "gbm_tcga_all")
head(df)
boxplot(df[,2] ~ df[,1], main="NF1 : CNA status vs mRNA expression", xlab="CNA status", ylab="mRNA expression", outpch = NA)
stripchart(df[,2] ~ df[,1], vertical=T, add=T, method="jitter",pch=1,col='red')
Alternatively, the generic cgdsr plot() function can be used to generate a similar plot:
In [ ]:
plot(mycgds, "gbm_tcga", "NF1", c("gbm_tcga_gistic","gbm_tcga_mrna"), "gbm_tcga_all", skin = 'disc_cont')
In this example, we evaluate the relationship of MDM2 and MDM4 expression levels in glioblastoma. mRNA expression levels of MDM2 and MDM4 have been median adjusted on the gene level (by globally subtracting the median expression level of the individual gene across all samples).
In [ ]:
df <- getProfileData(mycgds, c("MDM2","MDM4"), "gbm_tcga_mrna", "gbm_tcga_all")
head(df)
plot(df, main="MDM2 and MDM4 mRNA expression", xlab="MDM2 mRNA expression", ylab="MDM4 mRNA expression")
Alternatively, the generic cgdsr plot() function can be used to generate a similar plot:
In [ ]:
plot(mycgds, "gbm_tcga", c("MDM2","MDM4"), "gbm_tcga_mrna" ,"gbm_tcga_all")
In [ ]:
df.pri <- getProfileData(mycgds, "PTEN", "prad_mskcc_mrna", "prad_mskcc_primary")
head(df.pri)
df.met <- getProfileData(mycgds, "PTEN", "prad_mskcc_mrna", "prad_mskcc_mets")
head(df.met)
boxplot(list(t(df.pri),t(df.met)), main="PTEN expression in primary and metastatic tumors", xlab="Tumor type", ylab="PTEN mRNA expression",names=c('primary','metastatic'), outpch = NA)
stripchart(list(t(df.pri),t(df.met)), vertical=T, add=T, method="jitter",pch=1,col='red')