Skip to content
Tumor Covariate Signature Model
Python R
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
LUSC-SKCM
SomaticSignatures update simulated data results for paper Mar 21, 2019
TCGA-HR-experiment
TCGA-data
TCGA-dinucs
envs
experiments/simulated-data
mutation-signatures-viz @ cfc11ed
signature-estimation-py @ 3d0583d
signature-visualization
src
.gitmodules
.travis.yml
LICENSE
README.md
Snakefile

README.md

Tumor Covariate Signature Model (TCSM)

This repository contains the code for reproducing the results on real data from the paper Modeling Clinical and Molecular Covariates of Mutational Process Activity in Cancer, which implements the Tumor Covariate Signature Model (TCSM). The code for reproducing the simulated data results from the same paper are available upon request. TCSM requires Python 3. We recommend using Anaconda to install dependencies.

Once you have Anaconda installed, you can create an environment with the dependencies and activate it with the following commands:

conda env create -f environment.yml
conda activate tcsm

TCSM requires two submodules, mutation-signature-viz and signature-estimation-py, which can be installed by the following commands

git submodule init
git submodule update

References

Welles Robinson, Roded Sharan, Max Leiserson. (2019). Modeling clinical and molecular covariates of mutational process activity in cancer. Bioinformatics. paper link

Figures

Figure 3: (A) Comparison of the log-likelihood of held-out samples across K = 2–10 between TCSM with the biallelic HR covariate (inactivations of BRCA1, BRCA2 or RAD51C) and TCSM without covariates. (B) The log-likelihood ratio (LLR) of samples with the biallelic HR covariate hidden where LLR>0 indicates the mutations of a sample are more likely under the biallelic HR covariate inactivation model. (C) After excluding tumors with known biallelic inactivations in BRCA1, BRCA2 or RAD51C, the plot of a tumor’s LLR against its LST count

Homologous recombination repair (HR) deficiency in breast cancer

The main code for reproducing these experiments are in the TCGA-HR-experiment directory. From the top directory, use the following command to move into the TCGA-HR-experiment directory.

cd TCGA-HR-experiment

Reproducing Figure 3

snakemake figure_3

Reproducing Biallelic Inactivation Prediction

To calculate the AUPRC using the exposures estimated by TCSM with the biallelic covariate and TCSM without covariates

eval_HR_prediction_stm

To calculate the AUPRC using the exposures estimated by SomaticSignatures (should be in separate environment with SomaticSignatures installed)

eval_HR_prediction_SS

Reproducing Supplemental Figures S1-3

To reproduce figures S1 and S2

snakemake supplemental_figures_S1_S2

To reproduce figure S3

snakemake supplemental_figures_S3

Simultaneously learning signatures in melanomas and lung cancers

The main code for reproducing these experiments are in the LUSC-SKCM directory. From the top directory, use the following command to move into the LUSC-SKCM directory.

cd LUSC-SKCM

Reproducing Figure 4

snakemake figure_4

Supplemental Figures S4-5

snakemake supplemental_figures
You can’t perform that action at this time.