Ashford-A / PRECEPTS1

This repository houses a workflow that uses biological feature trees to segregate cancer RNA-seq datasets, then it trains machine learning models to predict the presence or absence of known, cancer-associated DNA-level mutations.

PRECEPTS1

The purpose of this project is to investigate whether we can use classification algorithms to not only predict the presence of different mutations recurrent in tumour cohorts, but also make biologically useful inferences about these mutations' downstream effects.

See experiments/subgrouping_test for the bulk of the code used in our publication based on this project, as well as a README explaining how to set up and run one of the experiment pipelines.

This repository consists of three major parts:

features

Collecting and processing expression, mutation, and other -omics datasets as well as phenotypes such as drug response.

predict

Custom machine learning tools to predict presence of gene mutations, drug response profiles, etc.

experiments

Scripts for running particular analyses.