gnomeR

the gnomeR package provides a consistent framework for genetic data processing, visualization and analysis. This is primarily targeted to IMPACT datasets but can also be applied to any genomic data provided by CbioPortal.

Dowloading and gathering data from CbioPortal through an integrated API using simply the sample IDs of the samples of interests or the name of the study to retrive all samples in that study. A separate package cbioportalr was developed independently.
Processing genomic data retrieved for mutations (MAF file), fusions (MAF file) and copy-number alterations (and when available segmentation files) into an analysis ready format.
Visualization of the processed data provided through MAF file summaries, OncoPrints and heatmaps.
Analyzing the processed data for association with binary, continuous and survival outcome. Including further visualiztion to improve understanding of the results.

Installation

You can install gnomeR from GitHub with:

# install.packages("devtools")
devtools::install_github("AxelitoMartin/gnomeR")

Similarly for those who wish to explore the development version of gnomeR:

devtools::install_github("AxelitoMartin/gnomeR", ref = "development")

Along with its companion package for cbioPortal data download:

devtools::install_github("karissawhiting/cbioportalr")

Examples

Setting up the API

In order to download the data from CbioPortal, one must first require a token from the website CbioPortal wich will prompt a login page with your MSKCC credentials. Then navigate to “Web API” in the top bar menu, following this simply download a token and copy it after running the following command in R:

usethis::edit_r_environ()

And pasting the token you were given in the .Renviron file that was created and saving after pasting your token.

CBIOPORTAL_TOKEN = 'YOUR_TOKEN'

You can test your connection using:

get_cbioportal_token()

Retrieving data

Now that the Cbioportal API is set up in your environment, you must first specify the database of interest (IMPACT or TCGA are the two available options). Following this one can either sepcify the samples or study of interest:

library(gnomeR)
library(cbioportalr)
ids <- as.character(unique(mut$Tumor_Sample_Barcode)[1:100])
df <- get_genetics(sample_ids = ids,database = "msk_impact",
                       mutations = TRUE, fusions = TRUE, cna = TRUE)

Processing the downloaded data

The binmat() function is the feature of the data processing of gnomeR. It takes genomic inputs from various sources of CbioPortal (mutation files, fusion files and copy number raw counts) to give out a clean binary matrix of n samples by all the events that were found in the files.

df.clean <- binmat(maf = df$mut, cna = df$cna)

We further included example datasets from the raw dowloaded files on CbioPortal (mut, fusion, cna) which we will use for the following examples.

set.seed(123)
patients <- as.character(unique(mut$Tumor_Sample_Barcode))[sample(1:length(unique(mut$Tumor_Sample_Barcode)), 100, replace=FALSE)]

gen.dat <- binmat(patients = patients, maf = mut, fusion = fusion, cna = cna)
kable(gen.dat[1:10,1:10],row.names = TRUE)

	TP53	KEAP1	KRAS
P-0010604-T01-IM5	1	0	0
P-0002651-T01-IM3	1	0	0
P-0000270-T01-IM3	1	0	0
P-0002915-T01-IM3	0	0	0
P-0011099-T01-IM5	0	0	0
P-0000080-T01-IM3	1	0	0
P-0001741-T01-IM3	1	1	1
P-0003964-T01-IM3	0	1	1
P-0003842-T01-IM5	0	0	0
P-0002597-T02-IM5	0	0	0

Visualization

MAF

Before we move on to more complex visualizations, we integrate the maf_viz() function to give an overview of the distribution of the different mutations across the cohort of interest:

sum.plots <- maf_viz(maf = mut %>% filter(Tumor_Sample_Barcode %in% patients))
sum.plots$topgenes

sum.plots$genecomut

OncoPrints

OncoPrints are a convenient way to display the overall genomic profiles of samples in the cohort of interest. This is best used for a subset of genes that are under consideration.

genes <- c("TP53","PIK3CA","KRAS","TERT","EGFR","FAT","ALK","CDKN2A","CDKN2B")
plot_oncoPrint(gen.dat = gen.dat %>% select(starts_with(genes)))

FACETs

FACETs is an ASCN tool and open-source software with a broad application to whole genome, whole-exome, as well as targeted panel sequencing platforms. It is a fully integrated stand-alone pipeline that includes sequencing BAM file post-processing, joint segmentation of total- and allele-specific read counts, and integer copy number calls corrected for tumor purity, ploidy and clonal heterogeneity, with comprehensive output.

p.heat <- facets.heatmap(seg = seg, patients = patients, min.purity = 0)
p.heat$p

Analysis

In this section we will quickly overview the possible analysis in gnomeR.

Binary and continuous outcomes

The gen.summary() function let’s the user perform a large scale association between the genomic features present in the binmat() function output and an outcome of choice:

binary (unpaired test using Fisher’s exact test and paired test using McNemmar’s exact test)
continuous (using simple linear regression)

outcome <- factor(rbinom(n = length(patients),size = 1,prob = 1/2),levels = c("0","1"))
out <- gen.summary(gen.dat = gen.dat,outcome = outcome,filter = 0.05)
kable(out$fits[1:10,],row.names = TRUE)

	Feature	Overall	0(N=50)	1(N=50)	OddsRatio	Pvalue	FDR	Lower	Upper
PIK3CA	PIK3CA	12%	8%	16%	2.17	3.57e-01	1.00e+00	0.53	10.6
EPHA5	EPHA5	5%	8%	2%	0.24	3.62e-01	1.00e+00	0	2.52
FLT4	FLT4	5%	2%	8%	4.21	3.62e-01	1.00e+00	0.4	213.8
DOT1L	DOT1L	5%	2%	8%	4.21	3.62e-01	1.00e+00	0.4	213.8
CDKN2B.Del	CDKN2B.Del	5%	2%	8%	4.21	3.62e-01	1.00e+00	0.4	213.8
STK11	STK11	7%	10%	4%	0.38	4.36e-01	1.00e+00	0.03	2.46
APC	APC	7%	4%	10%	2.64	4.36e-01	1.00e+00	0.41	29.07
MLL	MLL	9%	12%	6%	0.47	4.87e-01	1.00e+00	0.07	2.37
FAT1	FAT1	11%	14%	8%	0.54	5.25e-01	1.00e+00	0.11	2.29
KRAS	KRAS	13%	16%	10%	0.59	5.54e-01	1.00e+00	0.14	2.22

out$forest.plot

Survival analysis

Similarly we include simple tools to perform univariate Cox’s proportional regression adjusted for false discovery rate in the uni.cox() function.

time <- rexp(length(patients))
status <- outcome
surv.dat <- as.data.frame(cbind(time,status))
out <- uni.cox(X = gen.dat, surv.dat = surv.dat, surv.formula = Surv(time,status)~.,filter = 0.05)
kable(out$tab[1:10,],row.names = TRUE)

	Feature	Coefficient	HR	Pvalue	FDR	MutationFrequency
MLL	MLL	-1.48	0.23	0.0155	0.496	0.09
STK11	STK11	-1.46	0.23	0.0519	0.831	0.07
KEAP1	KEAP1	-1.02	0.36	0.1670	0.922	0.05
NOTCH1	NOTCH1	0.61	1.84	0.2070	0.922	0.08
DOT1L	DOT1L	0.66	1.93	0.2110	0.922	0.05
TSC1	TSC1	0.75	2.12	0.2130	0.922	0.05
CDH1	CDH1	0.60	1.83	0.2520	0.922	0.06
EPHA5	EPHA5	-1.08	0.34	0.2870	0.922	0.05
PIK3CA	PIK3CA	0.41	1.51	0.2950	0.922	0.12
PTPRD	PTPRD	0.49	1.63	0.3570	0.922	0.08

out$KM[[1]]

Further analytical tools

The primary goal of gnomeR not being in depth analysis of genomic data but rather reliable, modulable and reproducible framework for processing various types of genomic data. For users interested in large scale genomic analytical methods we compiled various packages developed by Department of Epidemiology and Biostatistics, Memorial Sloan-Kettering Cancer Center under an umbrella R package, gnomeVerse.

LurkingVariable/gnomeR