Evaluating the power and limitations of genome-wide association mapping in C. elegans

Abstract

A central goal of evolutionary genetics in Caenorhabditis elegans is to understand the genetic basis of traits that contribute to adaptation and fitness. Genome-wide association (GWA) mappings scan the genome for individual genetic variants that are significantly correlated with phenotypic variation in a population, or quantitative trait loci (QTL). GWA mappings are a popular choice for quantitative genetic analyses because the QTL that are discovered segregate in natural populations. Despite numerous successful mapping experiments, the empirical performance of GWA mappings has not, to date, been formally evaluated for this species. We developed an open-source GWA mapping pipeline called NemaScan and used a simulation-based approach to provide benchmarks of mapping performance among wild C. elegans strains. Simulated trait heritability and complexity determined the spectrum of QTL detected by GWA mappings. Power to detect smaller-effect QTL increased with the number of strains sampled from the C. elegans Natural Diversity Resource (CeNDR). Population structure was a major driver of variation in GWA mapping performance, with populations shaped by recent selection exhibiting significantly lower false discovery rates than populations composed of more divergent strains. We also recapitulated previous GWA mappings of experimentally validated quantitative trait variants. Our simulation-based evaluation of GWA performance provides the community with critical context for pursuing quantitative genetic studies using CeNDR to elucidate the genetic basis of complex traits in C. elegans natural populations.

Local adaptation and spatiotemporal patterns of genetic diversity revealed by repeated sampling of Caenorhabditis elegans across the Hawaiian Islands

“…We considered Bayes Factors (BFs) above 20 as evidence of a significant genotype-environment association. We also performed GEA with the NemaScan pipeline, a GWA tool specifically designed for C. elegans, which is available at https://github.com/AndersenLab/NemaScan (Widmayer, Evans, Zdraljevic, & Andersen, 2021).…”

“…However, methods to make GEA more robust to various types of structure exist. In the case of a GEA using GWA, the NemaScan pipeline accounts for the relatedness and genetic structure among strains with a genomic relationship matrix that can at least partially account for population structure (Widmayer et al, 2021;Yang et al, 2011). On the other hand, BayPass uses Bayesian hierarchical models to explicitly account for the scaled covariance matrix of population allele frequencies (Gautier, 2015).…”

Local adaptation and spatiotemporal patterns of genetic diversity revealed by repeated sampling of Caenorhabditis elegans across the Hawaiian Islands

“…We also explored associations with latitude and longitude because selective pressures associated with geographic location could also underlie signatures of local adaptation. Using the NemaScan GWA pipeline (Widmayer et al, 2021 ), we found 39 regions of the genome that were associated with environmental or geographic parameters (Figure 7 ). The BayPass genome‐wide scan revealed 108 regions of the genome that were associated with environmental or geographic parameters (Figure 7 ).…”

“…However, methods to make GEA more robust to various types of structure exist. In the case of a GEA using GWA, the NemaScan pipeline accounts for the relatedness and genetic structure among strains with a genomic relationship matrix (Widmayer et al, 2021 ; Yang et al, 2011 ). On the other hand, BayPass uses Bayesian hierarchical models to explicitly account for the scaled covariance matrix of population allele frequencies (Gautier, 2015 ).…”

“…However, in either approach, the inferences made when using these tools are subject to limitations. For example, simulations testing the performance of the NemaScan pipeline with various mapping panels indicate that false discovery rates are higher when strongly differentiated strains are included in the mapping panel rather than more closely related strains (Widmayer et al, 2021 ). This limitation means that some regions implicated in local adaptation might reflect unresolved population structure among the sample of Hawaiian strains.…”