2.4 | Genome-wide association analysis
To identify SNPs associated with virus exposure status we performed a
genome‐wide association study (GWAS) using BayPass 2.1 (Gautier 2015).
Analyses were performed under the auxiliary (AUX) covariate mode
(‐covmcmc and ‐auxmode flags), after scaling the variables with the
‐scalecov flag. The underlying models explicitly account for the
covariance structure among the population allele frequencies that
originates from the shared history of the populations through the
estimation of the population covariance matrix Ω, which removes the
variation associated with demography (Bonhomme et al., 2010; Gunther &
Coop, 2013). The auxiliary covariate model specifically involves the
introduction of a binary auxiliary variable to classify each locus as
associated or not associated. This allows computation of posterior
inclusion probabilities (and Bayes Factors) for each locus while
explicitly accounting for multiple testing issues. The auxiliary
covariate model was applied with default parameters, a 5,000 burn‐in of
iterations in the Markov chain Monte Carlo (MCMC) simulation, followed
by 25,000 iterations. To reduce artefacts due to potential variability
between runs, we performed 3 independent BayPass simulations. We then
calculated the average Bayes Factor (BF), expressed in deciban units
(dB), for each SNP as a quantitative estimate of the strength of
association with virus exposure and the standardized allele frequency.
For each SNP, the level of effect was assessed based on the Bayes Factor
(BF) models according to Jeffrey’s rule (Jeffreys 1961). SNPs with BF
scores ≥ 50 were regarded as decisive associations with virus exposure
and were retained as potential candidate loci.