genomic-features-survival

Scripts supporting identification of genomic features affecting survival time in cancer.

Getting Started

The easiest way to get started is to use requirements.txt to set up a conda environment with the relevant packages.

$ conda create --name <env> --file requirements.txt

The code in this repo was built with python 2.7, pandas 0.18, numpy 1.10, and scipy 1.0.0 (as is captured in requirements.txt)

The source data is stored in a public GCS bucket. Documentation for accessing public GCS data is here. The bucket for this project is called public-smith-sheltzer-cancer-analysis.

Use run.py to download the relevant data from the public GCS bucket and perform univariate analyses for copy number and mutation. run.py takes two optional arguments -- the folder to store data and analysis, (default .) and the number of parallel workers to use for the analysis (default none, and a sequential only analysis.). With -p 4, run.py takes ~12 hours on a 2017 MacBook Pro.

$ python run.py -p 4 -o $ouput_directory

Organization

• cbioportal - scripts used to analyze cbioportal data
• cnv-and-mutations - scripts for analyzing with cnas and and mutations together
• common-case-zscores - allows getting zscores for every row in a "common" file. common files have genes in rows, patients in columns.
• common - the common set of utilities and tools used in analysis.
  • analysis.py has the meat of cox analysis
  • mutation_base.py has the repeatable processing required to turn raw mutation data into usable dataframes.
• copy-number-analysis - given a copy number file, data about gene/location, and TCGA clinical data, calculate zscores for copy number genes
  • process_copy_numbers_to_genes.py has the repeatable processing to turn copy number raw data into usable dataframes.
• data-munging - contains scripts for miscellaneous small processing tasks: one-off zscores, density plot generation, etc
• fdr - scripts for performing false discovery correction
• geo - scripts for analyzing zscores for GEO files
• make-pancan - given a filetype/platform, take all the per-cancer-type zscore files and produce a file with genes in rows, and cancer types in columns
• mutation-analysis - given a tcga clinical file, and mutation data from the same set of patients, calculate zscores and kaplan meier curves for genes mutated in a sufficient number of patients
• pan-platform - scripts for creating panplatform TCGA files