parallelnewhybrid: an R package for the parallelization of hybrid detection using newhybrids

Abstract

Hybridization among populations and species is a central theme in many areas of biology, and the study of hybridization has direct applicability to testing hypotheses about evolution, speciation and genetic recombination, as well as having conservation, legal and regulatory implications. Yet, despite being a topic of considerable interest, the identification of hybrid individuals, and quantification of the (un)certainty surrounding the identifications, remains difficult. Unlike other programs that exist to identify hybrids based on genotypic information, newhybrids is able to assign individuals to specific hybrid classes (e.g. F , F ) because it makes use of patterns of gene inheritance within each locus, rather than just the proportions of gene inheritance within each individual. For each comparison and set of markers, multiple independent runs of each data set should be used to develop an estimate of the hybrid class assignment accuracy. The necessity of analysing multiple simulated data sets, constructed from large genomewide data sets, presents significant computational challenges. To address these challenges, we present parallelnewhybrid, an r package designed to decrease user burden when undertaking multiple newhybrids analyses. parallelnewhybrid does so by taking advantage of the parallel computational capabilities inherent in modern computers to efficiently and automatically execute separate newhybrids runs in parallel. We show that parallelization of analyses using this package affords users several-fold reductions in time over a traditional serial analysis. parallelnewhybrid consists of an example data set, a readme and three operating system-specific functions to execute parallel newhybrids analyses on each of a computer's c cores. parallelnewhybrid is freely available on the long-term software hosting site github (www.github.com/bwringe/parallelnewhybrid).

Extensive hybridization following a large escape of domesticated Atlantic salmon in the Northwest Atlantic

“…Next, using the centered baselines, individuals in generation t +1 were created by randomly sampling without replacement one allele per locus from each of the parental populations (i.e., wild baseline subsample and farmed baseline subsample) at time t 49 . Three independently simulated datasets were each in turn analyzed three times in parallel using NewHybrids 32 and the R package parallelnewhybrid 50 , with a burn-in of 50,000 followed by 100,000 sweeps. NEWHYBRIDS calculates the posterior probability that an individual belongs to each of, in our case, six hybrid classes 32 .…”

Estimating the relative fitness of escaped farmed salmon offspring in the wild and modelling the consequences of invasion for wild populations

“…We build directly on previous work which developed genetic and analytical tools to identify hybrids (Anderson & Thompson, 2002;Wringe, Stanley, Jeffery, Anderson, & Bradbury, 2017a, 2017b and documented interbreeding between escaped farmed and wild salmon following this escape event . We expand on these studies and others (Clifford, McGinnity, & Ferguson, 1998a, 1998bFleming et al, 2000;McGinnity et al, 2003;Skaala et al, 2012) by estimating the strength of selection against domestic and hybrid offspring in the wild, and explore the importance of obtaining accurate estimates of relative survival for predicting long-term consequences of invasion.…”

Habitat use and hybridisation between the Rocky Mountain sculpin (Cottussp.) and slimy sculpin (Cottus cognatus)

“…NEWHYBRIDS probabilistically classifies individuals based on their 10 locus genotypes into six genotypic classes: the two parental species, first (F1) and second (F2) generation hybrids and backcrosses to one or other parental species (BC 1 and BC2, respectively) (Anderson & Thompson, ). The NEWHYBRIDS model was built using the genepopedit , paralellnewhybrid , and hybriddetective packages in R software (Stanley, Jeffery, Wringe, Dibacco, & Bradbury, ; Wringe, Stanley, Jeffery, Anderson, & Bradbury, ,b). For our NEWHYBRIDS model, we also used a burn‐in of 50,000 generations and 500,000 MCMC generations to produce assignments of pure RMS, SS and hybrids across five simulated data sets.…”