GADGETS: a genetic algorithm for detecting epistasis using nuclear families

Abstract

Motivation: Epistasis may play an etiologic role in complex diseases, but research has been hindered because identification of interactions among sets of single nucleotide polymorphisms (SNPs) requires exploration of immense search spaces. Current approaches using nuclear families accommodate at most several hundred candidate SNPs.

Results: GADGETS detects epistatic SNP-sets by applying a genetic algorithm to case-parent or case-sibling data. To allow for multiple epistatic sets, island subpopulations of SNP-sets evolve separately under selection for evident joint relevance to disease risk. The software evaluates the identified SNP-sets via permutation testing and provides graphical visualization. GADGETS correctly identified epistatic SNP-sets in realistically simulated case-parent triads with 10 000 candidate SNPs, far more SNPs than competitors can handle, and it outperformed competitors in simulations with many fewer SNPs. Applying GADGETS to family-based oral-clefting data from dbGaP identified SNP-sets with possible epistatic effects on risk.

Availability and implementation: GADGETS is part of the epistasisGA package at https://github.com/mnodzenski/epistasisGA.

Supplementary information: Supplementary data are available at Bioinformatics online.

Fig. 1.

Flowchart of the GADGETS algorithm for a single island. In this simple illustration, we include four chromosomes of size d = 3. The text describes chromosome migration among islands

Fig. 2.

Network plot for scenario 1, replicate 2. Chromosomes were filtered for inclusion using global permutations. SNP labels ‘1’ and ‘2’ indicate membership in epistatic risk sets 1 and 2, respectively. Larger, darker nodes and thicker, darker edges correspond to larger SNP and SNP-pair scores, respectively. Dashed connections indicate pairs of SNPs located on the same biological chromosome with pairwise R² of at least 0.1 in complement pseudo-siblings

Fig. 3.

Network plot for Asian (cleft lip with or without cleft palate) case-parent triad data. Chromosomes were filtered for inclusion using global permutations. Larger, darker nodes and thicker, darker edges correspond to larger SNP and SNP-pair scores. Dashed connections indicate pairs of SNPs located on the same biological chromosome with pairwise R² of at least 0.1 in complement pseudo-siblings

Fig. 4.

Network plot for European (cleft lip with or without cleft palate) case-parent triad data. Chromosomes were filtered for inclusion using global permutations. Larger, darker nodes and thicker, darker edges correspond to larger SNP and SNP-pair scores. Dashed connections indicate pairs of SNPs located on the same biological chromosome with pairwise R² of at least 0.1 in complement pseudo-siblings