Potpourri provides a MATLAB interface for ease of use. These instructions will guide you to build and run Potpourri on MATLAB.
Building Potpourri requires the Boost C++ library.
In order to build Potpourri for MATLAB, just type make on the terminal!:
make
or directly run the MATLAB script for building mex files:
build_mex.m
@ Feature Matrix:
This should consist of a grid {0, 1, 2} characters, representing homozygous major, heterozygous and homozygous
minor genotypes respectively for all samples. Each row corresponds to a sample.
@ Labels:
This should consist of {0, 1} binary labels representing control and case respectively.
@ SNP Information:
This should consist of three columns: unique SNP_id, chromosome and position
@ Regulatory/Coding Information:
This should consist of {0, 1} binary labels representing whether the corresponding SNP (with columns of feature matrix) is in the regulatory region (1) or not (0).
@ Network Matrix:
An adjacency matrix for SNP-SNP interaction.
@ Maximum marginal significance:
Takes an integer value from 1-6, representing the maximum marginal significance of loci for consideration in pairwise testing as a -log10(p-value).
@ outputFileName:
Prefix for output files.
@ Omega:
A float parameter of Potpourri to reward regulatory region.
@ b:
Number of neighbors that should be included in the epistasis test for each selected SNP.
@ k:
Number of features to be selected
How to run Potpourri on MATLAB. Simply run the demo file:
demo_potpourri.m
The example data is adapted from Atwell et. al. (2010). The genotype and phenotype data of Arabidopsis Thaliana (AT) obtained from Atwell et. al. (2010) and adapted to the algorithm accordingly. For descriptions and format of the data, check the readme file for data.
After running Potpourri, files above will be either created, or appended to, in the output directory. ###.Summary The summary file contains several statistics such as statistical tests performed, number of reciprocal pairs found, Each row corresponds to a separate run. ###.Cutoff Pairs Contains all detected pairs above the dynamic significance threshold at the conclusion of a run. ###.Reciprocal Pairs Contains a subset of the pairs in the cutoff pairs set such that both loci in a pairing had the other locus selected as its most significant interaction. ###.Reciprocal Pairs Formatted Each row represents a single reciprocal locus pairing with chi-squared significance, locus 1 and 2, chromosome 1 and 2, base pair 1 and 2.
This project is licensed under GNU GPL v3 - see the LICENSE.md file for details.
We used LINDEN and SPADIS code in our implementation from the studies by Cowman et. al. and Yilmaz et. al respectively.
Yilmaz, Serhan, Tastan, Oznur & Cicek, A. Ercument (2018). SPADIS: An Algorithm for Selecting Predictive and Diverse SNPs in GWAS. bioRxiv
Atwell, S. et al. (2010) Genome-wide association study of 107 phenotypes in Arabidopsis thaliana inbred lines. Nature, 465(7298), 627–631.
Wu, M. C. et al. (2011) Rare-variant association testing for sequencing data with the sequence kernel association test. The American Journal of Human Genetics, 89(1), 82–93.
Cowman T, Koyutürk M. (2017) Prioritizing tests of epistasis through hierarchical representation of genomic redundancies. Nucleic acids research, 45(14), e131.