STRUCTURE and ROLLOFF reimplementations for CS 4775 (Computational Genetics and Genomics) at Cornell

Original STRUCTURE paper: "Inference of Population Structure Using Multilocus Genotype Data"

Original ROLLOFF paper: "The History of African Gene Flow into Southern Europeans, Levantines, and Jews"

To install in a virtual environment:

python3 -m venv env
source env/bin/activate
pip install -r requirements.txt

structure.py infers admixture proportions given a dataset of genetic variants in Variant Call Format. The output is written to an HDF5 file, which is used by visualize.py and rolloff.py.

python structure.py [-h] [-k num_populations] [-o output.hdf5]
                    [--profile] [-d drop_frac] [-m num_burn_in_rounds]
                    [-s num_samples] [-c num_rounds_btwn_samples] data_file

Command-line arguments:

• data_file: the input file, either a VCF file or an Eigenstrat (.phgeno) file
• -k: the number of populations
• -o or --out: the HDF5 file to which the output will be written
• -d or --drop-frac: the fraction of loci to drop
• -m or --burn-in: the burn-in period
• -s or --num-samples: the number of samples to collect
• -c or --sample-interval: number of rounds between samples

structure.py can also take a .phgeno data file as its primary argument.

visualize.py creates charts to display the output from structure.py. It takes one command-line argument, the location of the HDF5 file.

rolloff.py estimates the ROLLOFF statistic for two populations. It takes two command-line arguments:

rolloff.py [-h] [--profile] [-m centimorgans] data_file.hdf5

• data_file.hdf5: the file generated by structure.py
• -m or --min-bin-size: the minimum bin size to use, in centimorgans

The data files are compressed and stored in the data/ folder. They are from the 1000 Genomes repository. Use of these datasets is subject to these terms.

• data/1.1-200000.ALL.chr1_GRCh38.genotypes.20170504.vcf.gz (the 200K dataset) contains variant calls from the first 200,000 positions in chromosome 1.
• data/1.1-2500000.ALL.chr1.phase3_shapeit2_mvncall_integrated_v5a.20130502.genotypes.vcf.gz (the 2.5M dataset) contains variant calls from the first 2.5 million positions in chromosome 1.