Statistical inference using IBD segments

isweep is a Python package and a series of automated workflows to study natural selection with identity-by-descent (IBD) segments. The Python package simulates IBD segments around a locus and estimates selection coefficients. The automated workflows perform selection scans, selection coefficient estimation, IBD case-control mapping, haplotype phasing, and local ancestry inference. Scripts in the workflows can be run individually in scripts/, with argparse documentation and inputs.

These methods are suitable for analyses for recent genetic/evolution events. For example,

• By recent, we mean within the last 500 generations.
• By strong, we mean selection coefficient s >= 0.015 (1.5%).
• Scan may have moderate power for s >= 0.01 (1%).

Please review the Readthedocs for detailed support, including which relevant publications to cite if you use this software.

Please file an Issue on GitHub for troubleshooting.

Contact sethtem@umich.edu for support specific to your analysis, e.g., analyses of non-human genetic data.

The input data is:

• Whole genome sequences
  • Probably at least > 500 diploids
  • Phased vcf data 0|1 of recombining chromosomes
  • Tab-separated genetic map (bp ---> cM)
    • Without headers!
    • Columns are chromosome, rsID, cM, bp
• Access to cluster computing

workflow/phasing-ancestry provides support for phasing and selecting an ancestry cohort.

Primary pipelines:

The main workflows, workflow/scan-selection and workflow/model-selection do:

1. Scan genome for extreme IBD rates
2. Detect anomalously large IBD clusters
3. Rank alleles based on evidence for selection
4. Compute a cluster agglomeration measure
5. Estimate frequency, location of unknown sweeping allele
6. Estimate a selection coefficient (w/ CIs)

In general, you run workflows with

nohup snakemake -s Snakefile-*.smk -c1 --cluster "sbatch [options]" [options] --jobs XX --configfile *.yaml &

You modify the relevant YAML files, which define the method parameters. You should run the pipelines in the mamba activate isweep environment.

Step 1 may be standalone, depending on the analysis. (You may not care to model putative sweeps (Steps 2-6), which also requires demographic Ne estimation.)

Installation

To install the dependencies and our package:

1. Clone the repository

git clone https://github.com/sdtemple/isweep.git 

2. Get the Python package

mamba env create -f isweep-environment.yml

3. Download some Java software.

bash get-software.sh 

You can test the workflows with our small Zenodo repository.

Picture of selection scan

The flow chart below shows the steps ("rules") in the selection scan pipeline.