/SISTR

Selection Inference on Short Tandem Repeats (SISTR): A method to measure negative selection at short tandem repeats

Primary LanguagePython

SISTR: Selection Inference on Short Tandem Repeats

SISTR is a method to measure negative selection at short tandem repeats.

It takes allele frequency data (per-locus frequencies of each allele length) as input and outputs a posterior estimate of the selection coefficient at each locus.

For questions on SISTR usage or setup, please contact Bonnie Huang (bbhuang@ucsd.edu)

Installation

SISTR uses Python3 and the following libraries in addition to the Python Standard Library: SciPy, NumPy

You can obtain SISTR by cloning the Github repository:

git clone https://github.com/BonnieCSE/SISTR
cd SISTR

Pipeline

  1. The first step is a preprocessing step which involves simulating allele frequencies to generate lookup tables used later in step 2. Due to the compute time required to generate the lookup tables, we recommend using the following precomputed tables found here by downloading the entire folder sistr_resources/ and saving it in the same directory where you are running SISTR. However, if you wish to create your own custom lookup tables, see simulations/ for further details.

    Example command to run simulations to generate a custom ABC lookup table:

    python ./simulations/ABC_lookup.py \
  --out-folder sistr_resources_test/abc_lookup/ \
  --per 3 \
  --opt-allele 5 \
  --num-sims 5

  2. The second step is to run SISTR, which requires (1) allele frequency data and (2) the lookup tables generated in step 1 as input. It outputs a posterior estimate of the selection coefficient at each locus.

    See sistr/ for further details.

    Example command to run SISTR:

    python sistr/SISTR_v1.py \
  --in-file sistr/allele_freqs_test.txt \
  --out-file test_results.txt

Folder Contents

  • simulations/: Contains scripts to run simulations to generate lookup tables
  • sistr/: Contains scripts to run SISTR

FAQ

Q: If I decide to create my own lookup tables, what are the main features I can customize ?

A: Currently, for the ABC lookup tables, you can customize the number of simulations, the number of generations the simulations are run for, and the a (shape) and b (scale) parameters of the prior gamma distrbution from which s is drawn.
For the LRT lookup tables, you can customize the number of simulations as well as the number of generations the simulations are run for.