AF-vapeR: A multivariate genome scan for detecting parallel evolution using allele frequency change vectors

Abstract

The repeatability of evolution at the genetic level has been demonstrated to vary along a continuum from complete parallelism to divergence. In order to better understand why this continuum exists within and among systems, hypotheses must be tested using high-confidence sets of candidate loci for repeatability. Despite this, few methods have been developed to scan SNP data for signatures specifically associated with repeatability, as opposed to local adaptation. Here we present AF-vapeR (Allele Frequency Vector Analysis of Parallel Evolutionary Responses), an approach designed to identify genome regions exhibiting highly correlated allele frequency changes within haplotypes and among replicated allele frequency change vectors. The method divides the genome into windows of an equivalent number of SNPs, and within each window performs eigen decomposition over normalised allele frequency change vectors (AFV), each derived from a replicated pair of populations/species. Properties of the resulting eigenvalue distribution can be used to compare regions of the genome for those exhibiting strong parallelism, and can also be compared against a null distribution derived from randomly permuted AFV. We demonstrate the utility of this approach to detect different modes of parallel evolution using simulations, and also demonstrate a reduction in error rate compared with intersecting FST outliers. Lastly, we apply AF-vapeR to three previously published datasets (stickleback, guppies, and Galapagos finches) which comprise a range of sampling and sequencing strategies, and lineage ages. We highlight known parallel regions whilst also identifying novel candidates. The main benefits of this approach include a reduced false-negative rate under many conditions, an emphasis on signals associated specifically with repeatable evolution as opposed to local adaptation, and an opportunity to identify different modes of parallel evolution at the first instance.

Rapid genomic convergent evolution in experimental populations of Trinidadian guppies (Poecilia reticulata)

“…Further, applying eigenanalysis (AF‐vapeR) to the natural population dataset reveals that this region is an outlier on eigenvector 2, indicating multiple axes of HP to LP AF change across northern Trinidad (Whiting et al. 2021b ). Biologically, this is consistent with different but functionally similar variants.…”

“…The method is fully described and validated with simulated data in Whiting et al. ( 2021b ). We focused only on the first eigenvector here to highlight cases where all four populations were evolving in parallel.…”

“…The code to perform this method is available as an R package: AF-vapeR (https://github.com/JimWhiting91/afvaper). The method is fully described and validated with simulated data in Whiting et al (2021b). We focused only on the first eigenvector here to highlight cases where all four populations were evolving in parallel.…”

Selection-driven trait loss in independently evolved cavefish populations

“…Recent advances in analytical tools provide unprecedented opportunity to compare genome-wide patterns of genetic variation across lineages that have independently evolved similar phenotypes in response to similar environmental pressures. These powerful approaches can be used to infer the genetic and mechanistic basis of adaptive trait evolution and to better understand the predictability of evolution in natural populations [7][8][9] . Identifying the molecular basis of repeated evolution has traditionally posed a major challenge [10][11][12] .…”

“…Despite decades of research on the biological mechanisms underlying traits in cavefish, the drivers of repeated cave evolution remain unclear. Here we leverage large-scale whole genome sequencing of nearly 250 A. mexicanus individuals combined with cutting-edge population genomic methodologies 7,8,36,37 to quantify selection across the genome in multiple cave and surface populations to investigate how often we might expect repeated evolution of the same phenotype to take repeated paths at the molecular level. We also ask whether any commonalities exist among genes that are the repeated target of evolution.…”