This repository is deprecated please see the new MTS pipeline

Prism Data Processing

Public version of the data processing pipeline for PRISM medium throughput screens (MTS). For use by collaborators to regenerate tables and plots correlating drug response to cell line features. Public cell line data for equivalent analysis is available on the DepMap Portal.

For more general biomarker analysis see the cdsr_biomarker package which is based on models in the cdsr_models package.

This repository contains 3 primary scripts:

1. make_logMFI.R
2. MTS_Data_Processing.R
3. MTS_Analysis.R

FIRST run setup.R either in RStudio or terminal to install packages required for analysis. For shell execute:

cd ./prism_data_processing
Rscript src/setup.R

For more information and FAQs see the info folder.

make_logMFI.R

Only necessary for processing raw files downloaded from clue.io.

Converts raw .gctx and .txt files downloaded from clue.io to readable logMFI.csv. This file contains raw log2 median fluorescence intensity (MFI) data for each cell line at each treatment. Requires:

• PR300_LMFI.gctx: readout for PR300
• PR500_LMFI.gctx: readout for PR500
• PR300_inst_info.txt: treatment info for PR300
• PR300_cell_info.txt: cell line info for PR300
• PR500_inst_info.txt: treatment info for PR500
• PR500_cell_info.txt: cell line info for PR500
• skipped_wells.csv: file indicating which wells did not receive compound (optional)

MTS_Data_Processing.R

Does pre-processing based on raw median fluorescent intensity (MFI) values and generates tables with data on QC, viabilities, and dose-response parameters, as well as figures showing dose response curves.

Steps of pre-processing outlined in MTS_pipeline.md.

MTS_Analysis.R

Generates biomarker analysis for the processed data, including, univariate and multivariate analyses. Requires a directory of expression data (RNA, mutations, etc.) and the results of MTS_Data_Processing.R. See below for more details on each analysis function. Relies on analysis_functions.R.

NOTE: in order to run biomarker analysis, files containing omics data for cell lines must be downloaded from DepMap. In particular we recommend the latest versions of:

• Achilles_gene_effect.csv (CRISPR dependencies)
• CCLE_expression.csv (gene expression)
• primary-screen-replicate-collapsed-logfold-change.csv (repurposing)
• D2_Achilles_gene_dep_scores.csv (shRNA)
• CCLE_metabolomics_20190502.csv (metablomics)
• CCLE_RPPA_20181003.csv (proteomics)
• CCLE_miRNA_20181103.gct (miRNA)
• CCLE_gene_cn.csv (copy number)
• CCLE_mutations.csv (mutations)
• sample_info.csv (lineages)

Once these files are downloaded, use biomarker_tables.R to convert tables to matrix form (this may require tweaking due to changes in file structures). This R script also generates two combined datasets: x-ccle.csv and x-all.csv. These are used for multivariate models and are based on CCLE data and all DepMap data respectively.

MTS_functions.R and analysis_functions.R

These files contain helper functions used in the scripts above and are sourced at the beginning of each (to install necessary packages and define functions). analysis_functions.R contains functions used in MTS_Analysis.R to generate biomarker analyses, while MTS_functions.R contains functions used in MTS_Data_Processing.R.

In analysis_functions.R, each function takes a matrix of features (X) and a vector of responses (y) as input. multivariate fits both an elastic net and random forest to the data.

---

This repository is maintained by Cancer Data Science at the Broad Institute of MIT and Harvard