broadinstitute/cdsr_biomarker

R toolkit for biomarker analysis

cdsrbiomarker

cdsrbiomarker is an R toolkit for biomarker analysis. It includes helpful functions and standard reports.

Install

library(devtools)
devtools::install_github("broadinstitute/cdsr_models")
devtools::install_github("broadinstitute/cdsr_biomarker")

The package can then be loaded by calling

library(cdsrbiomarker)

Taiga

Many of the functions in the package use data which is stored on taiga. If you are at the Broad you can install taigr the taiga client for R by following the instructions here.

Functions

get_biomarkers

get_biomarkers is a wrapper around the discrete_test, lin_associations, and random_forest functions in cdsrmodels. It takes a response vector y and then loads features matricies from Taiga. The appropriate modeling functions are applied to each feature matrix and the results are returned and saved in the out_path directory if it is specified.

biomarker_results <- get_biomarkers(y, p_cutoff=0.1, out_path="~/Desktop/example")

Reports

The reports directory contains standard biomarker reports.

• single_profile_biomarker_report generates a biomarker report for a single response profile.

• multi_profile_biomarker_report compares biomarkers for multiple response profiles e.g drug and genetic or multiple drugs.

There are wrapper functions in cdsrbiomarker to automaticaly genenarate these reports. Here is an example using Achilles data for EGFR and PRISM data for a few EGFR inhibitors:

1. Make a cell line by perturbation response matrix Y.

gene_effect <- load.from.taiga(data.name='depmap-a0ab',data.file='Achilles_gene_effect')[,"EGFR (1956)"] %>% 
  enframe(name = "arxspan_id",value = "xpr_egfr")
auc <- load.from.taiga(data.name='secondary-screen-15e6', data.file='secondary_merged_drc_parameters') %>% 
  filter(repurposing_name %in% c("erlotinib","gefitinib","lapatinib")) %>% 
  select(auc,repurposing_name,arxspan_id) %>% 
  spread(key = "repurposing_name",value = "auc")
Y <- full_join(gene_effect,auc, by = "arxspan_id") %>% column_to_rownames(var = "arxspan_id") %>% as.matrix()

corner(Y)

##               xpr_egfr erlotinib gefitinib lapatinib
## ACH-000004  0.16303435        NA        NA        NA
## ACH-000005  0.05234623        NA        NA        NA
## ACH-000007 -0.27242329 1.2249556 0.6756661 0.6937505
## ACH-000009 -0.78660538        NA        NA        NA
## ACH-000011 -0.10197377 0.7143025 0.7616728 0.7627777

2. Make a meta data table which will be displayed in the report.

meta_data <- list(perturbation = colnames(Y), type = c("CRISPR","Drug","Drug","Drug")) %>% as_tibble()
meta_data

## # A tibble: 4 x 2
##   perturbation type  
##   <chr>        <chr> 
## 1 xpr_egfr     CRISPR
## 2 erlotinib    Drug  
## 3 gefitinib    Drug  
## 4 lapatinib    Drug

3. Call the generate report function for the report you want. You will need to give it a file path to save the results to.

cdsrbiomarker::generate_multi_profile_biomarker_report("~/Desktop/example/","example_title",Y,meta_data)

4. If you already have the biomarker results files and just want to generate the report you can do it like this.

cdsrbiomarker::generate_multi_profile_biomarker_report("~/Desktop/example","example")