raphael_examples/pancan/README.md

In jhrcook/wext: Weighted Exclusivity Test (WExT)

Running WExT WRE on Pan-Cancer data (pancan)

Description

This simple example generates two datasets -- which can be interpreted as data from two cancer types -- and then runs the WExT weighted row exclusivity (WRE) test on the combined, pan-cancer dataset in two modes. These two modes were inspired by Y.A. Kim, et al. (ISMB/Bioinformatics, 2015).

In the first mode, we are looking for pan-cancer exclusivity where we are controlling for cancer type. In other words, we don't want to find exclusivity that is driven by cancer type, i.e. where one gene is only mutated in cancer type A and a second gene is mutated only in cancer type B. Kim, et al. call this across mutual exclusivity ("ACROSS_ME"), although their ACROSS_ME also requires that the exclusivity isn't present in a single cancer type.

To run WExT in this mode, we process mutations and generate mutation probabilities for each dataset/cancer type separately, and then combine these when computing the WRE test across both datasets.

In the second mode, we are looking for pan-cancer exclusivity without controlling for cancer type. This is called between mutual exclusivity ("BETWEEN_ME") by Kim, et al. To perform this type of run in WExT, the datasets need to be combined

Data

To demonstrate these two modes, we generate simulated data. The data includes two implanted gene sets mutated in approximately 25% of samples. (G1, G2): this set has "across" mutual exclusivity, i.e. exclusivity where both genes are mutated in both datasets/cancer types. (G3, G4): this set has "between" mutual exclusivity, i.e. each gene is mutated in only one dataset/cancer type.

The datasets also include other genes that are mutated in each sample at a fixed background mutation rate. We ensure that every sample has at least one mutation.

Usage

Run make to process the mutations, compute mutation probabilities, and run the WRE test in both modes described above. You can find the two output TSV files in output/.

You can pass in a few optional parameters to make, including: K: gene set size (default: 2) NP: number of permutations (default: 100) NUM_GENES: number of genes in simulated dataset (default: 50) NUM_SAMPLES: number of samples in simulated dataset (default: 100) RANDOM_SEED: random seed for PRNG (default: 2) BMR: background mutation rate (default: 0.01)

For example, to run the example searching for sets of size three with 100 permutations:

make K=3 NP=100

Both these examples -- using K=2 or K=3 -- take less than 5 minutes when running on a single core on a modern machine.

The easiest way to interpret the results is to compare the set P-value (G3, G4) with between exclusivity in the two output files in output/. The P-value for (G3, G4) should be smaller (more significant) in the "between" output file than the P-value for (G3, G4) in the "across" output file.

jhrcook/wext documentation built on May 17, 2021, 1:19 a.m.