This repository contains code to run the Combined Analysis of Epistasis and Pleiotropy, or CAPE. This method, described in [1,2] combines information across multiple quantitative traits to infer directed genetic interactions. By combining information across multiple traits, CAPE not only increases power to detect genetic interactions, but also interprets these interactions across traits to identify a single interaction that is consistent across all observed data. CAPE can be applied to a variety of genetic variants, such as single nucleotide polymorphisms (SNPs), copy number variations (CNVs) or structural variations (SVs). Included here is also a small example data set with code to infer a predictive network between quantitative trait loci (QTL) in a BXD mouse population assayed for three immune phenotypes. We used phenotypes 16320, 10062, and 13011 from GeneNetwork (www.genenetwork.org). In this example, CAPE generates a genetic interaction network that describes how variants interact with each other to influence this group of related traits.
The traits used here are as follows:
-
16320 Immune system: ELISA-3x, IgG class antibody binding to TSHR A-subunit protein in ELISA 4 weeks after 3 immunizations with TSHR A-subunit adenovirus [OD490 nm]
-
10062 Immune system, pulmonary system: Tumor necrosis factor (TNF)-alpha level in lung after aerosolized lipopolysaccharide (LPS) exposure [pg/ml]
-
13011 Infectious disease, immune function: H1N1 (PR8) influenza A virus (2x10E3 FFU), median body weight loss day 7 after infection in 9-14 week-old females [%]
The parameters for CAPE are supplied with a parameter file. An example is supplied to show the basic format of the file. The parameters specified through the parameter file are as follows:
- Traits Names of traits to be analyzed
- scan.what Whether to analyze eigentraits or original traits
- traits.scaled Wether to scale traits
- traits.normalized Whether to normalize trait
- eig.which If eigentraits are being analyzed, which of the eigentraits to use
- pval.correction What type of p value correction method to use
- use.kinship Wether a kinship correction should be implemented
- kinship.type What type of kinship correction to implement. Either "overall" or "LTCO"
- pop What type of population is being analyzed: 2PP = two-parent, MPP - multi-parent, RIL = recombinant inbred lines
- ref.allele Which allele to use as the reference allele. Typically this is A, but may be different in a multi-parent population.
- singlescan.perm How many permutations to run for the single-locus scan. This is only to test single-locus significance for the user. CAPE does not consider significance when selecting markers for the pair scan.
- marker.selection.method The marker selection method to use. Typically top.effects. But from.file can also be used to specify specific markers to test.
- peak.density When marker.selection.method is top.effects, this parameter indicates how densely to sample markers under effect size peaks. A value of 0.5 indicates that half the markers under an effect size peak will be selected for pairwise testing.
- tolerance How many markers away from the specified number (num.alelles.in.pairscan) can be tolerated.
- num.alleles.in.pairscan How many markers should be selected for pairwise testing.
- max.pair.cor The maximum correlation between markers for pairwise testing. Testing markers that are highly correlated may lead to false positives.
- pairscan.null.size The size of the null distribution to generate for significance testing in the pairwise scan. We recommend at least 500k for significance testing.
-
Carter, G. W., Hays, M., Sherman, A. & Galitski, T. Use of pleiotropy to model genetic interactions in a population. PLoS Genet. 8, e1003010 (2012).
-
Tyler, A. L., Lu, W., Hendrick, J. J., Philip, V. M. & Carter, G. W. CAPE: an R package for combined analysis of pleiotropy and epistasis. PLoS Comput. Biol. 9, e1003270 (2013).