BayesSUR-RE

This repository is for the results in the paper Multivariate Bayesian structured variable selection for pharmacogenomic studies by 'Zhao Z., Banterle M., Lewin A. and Zucknick M. (2022)', arXiv:2101.05899v3.

Data

Abstract

The main data is the Genomics of Drug Sensitivity in Cancer (Garnett et al., 2012, Nature) which is publicly available. The gene set of MAPK is also is publicly available from Kyoto Encyclopedia of Genes and Genomes (KEGG).

Availability

No restrictions.

Description

The pharmacogenomic datasets in Garnett et al. (2012, Nature) can be downloaded from

ftp://ftp.sanger.ac.uk/pub4/cancerrxgene/releases/release-5.0/

There are three data files needed: gdsc_en_input_w5.csv, gdsc_drug_sensitivity_fitted_data_w5.csv and gdsc_tissue_output_w5.csv.

The gene set of MAPK can be downloaded from KEGG gene set enrichment analysis

http://software.broadinstitute.org/gsea/msigdb/cards/KEGG_MAPK_SIGNALING_PATHWAY

Download the .txt version and delete the head content, i.e. the first two lines.

Code

Abstract

We are including all of the code that will enable reproducing our results.

Description

We developed an R package BayesSUR which is available on CRAN for implementing our approach. All of the code are also available on the first author’s GitHub.

Optional Information

The real data analysis is computationally intensive due to high-dimensional genomic predictors. Running with one thread of CPU takes a few hours, but producing the same results in this article. It can be run faster on a cluster with parallelization. However, since the core code of our approach is in C++ for computational efficiency, Rcpp with OpenMP for parallelization is difficult reproduce the results on different type of machines.

Instructions for Use

Reproducibility

All tables, Figures 3-5, and Figures 7-10 can be reproduced through the provided code. The general steps are:

1. Load simulation function through file simulation_function.R.
2. Run script simulation_results.R line by line to reproduce all simulation results in the article.
3. Download real data by following the above Data section.
4. Run script GDSC_preprocess1.R to get the real data ready for modelling (need to uncomment some lines to obtain datasets corresponding 'Feature sets II and III' in the article); run script GDSC_preprocess2.R to get the independent validation data.
5. Run script GDSC_results.R line by line to reproduce all results of real data analysis in the article.