edawson / assisi Goto Github PK

Eric T. Dawson
September 27 2017

assisi is a python package that simulates mutational signature profiles (i.e. counts / proportions of trinucleotide context mutations) using Monte Carlo simulation.

We've included the cosmic mutational signatures probabilities from deconstructSigs (https://github.com/raerose01/deconstructSigs).

python assisi.py -h for help.

The following command will generate a single mutational catalog with 1000 total mutations. -r specifies that assisi should generate a random number of signatures (1-10) and assign random proportions to each of those ten signatures. The sum of the proportions is 1.0, with some error.
-i is the COSMIC SBS96 mutational context probabilities; each row is a signature, and each column is the proportion of that mutational context in the signature.

python assisi.py -i cosmic_file.tsv -n 1000 -r > catalog.tsv

Rules for COSMIC sig files:
No headers
Tab separated
No restriction on number of rows
There must be 96 columns (for now)

The next version of assisi will use a tidy format for this data.

The output of assisi when run as above is a single line with the following columns:

<number of signatures> <Python dictionary of signature : proportion> <first context count> ... <last context count>

Multiple catalogs can be generated by passing the -m flag and a desired number of catalogs:

## Simulate 1000 catalogs
python assisi.py -i cosmic_file.tsv -n 1000 -r -m 1000 > catalogs.tsv

Assisi uses the Python multiprocessing package to optionally generate catalogs in parallel. We recommend using this feature when simulating more than one catalog, as it decreases runtime proportional to the number of available compute cores and does not affect results in any way.

## Simulate 1000 catalogs, with 4 threads
python assisi.py -i cosmic_file.tsv -n 1000 -r -m 1000 -t 4 > catalogs.tsv

assisi is written in barebones python3 and is backwards-compatible with python 2.7 - no installation necessary.