Substantial differences in bias between single‐digest and double‐digest RAD‐seq libraries: A case study

Abstract

The trade-offs of using single-digest vs. double-digest restriction site-associated DNA sequencing (RAD-seq) protocols have been widely discussed. However, no direct empirical comparisons of the two methods have been conducted. Here, we sampled a single population of Gulf pipefish (Syngnathus scovelli) and genotyped 444 individuals using RAD-seq. Sixty individuals were subjected to single-digest RAD-seq (sdRAD-seq), and the remaining 384 individuals were genotyped using a double-digest RAD-seq (ddRAD-seq) protocol. We analysed the resulting Illumina sequencing data and compared the two genotyping methods when reads were analysed either together or separately. Coverage statistics, observed heterozygosity, and allele frequencies differed significantly between the two protocols, as did the results of selection components analysis. We also performed an in silico digestion of the Gulf pipefish genome and modelled five major sources of bias: PCR duplicates, polymorphic restriction sites, shearing bias, asymmetric sampling (i.e., genotyping fewer individuals with sdRAD-seq than with ddRAD-seq) and higher major allele frequencies. This combination of approaches allowed us to determine that polymorphic restriction sites, an asymmetric sampling scheme, mean allele frequencies and to some extent PCR duplicates all contribute to different estimates of allele frequencies between samples genotyped using sdRAD-seq versus ddRAD-seq. Our finding that sdRAD-seq and ddRAD-seq can result in different allele frequencies has implications for comparisons across studies and techniques that endeavour to identify genomewide signatures of evolutionary processes in natural populations.

The future of parentage analysis: From microsatellites to SNPs and beyond

“…Some important issues when choosing software for parentage analysis for next‐generation markers, beyond whether they can analyse the type of parentage data collected (e.g., parent–offspring pairs, groups of putative siblings and putative parents, or parent–parent–offspring triads), are (a) whether the program can handle the number of markers used in the study and (b) whether the method can accept genotype likelihoods that reflect the genotype uncertainties characteristic of next‐generation sequencing or whether additional consideration of errors will be required. All of the methods worth mentioning incorporate error rates, but most of those error rates are based on expectations for microsatellites and will likely not properly incorporate error arising from sequencing errors, allelic dropout and PCR bias, all of which can dramatically impact genotypes in next‐generation sequencing data sets such as RAD‐seq data (Flanagan & Jones, ).…”

“…Rather, the advice is to remove loci suffering from null alleles or allelic dropout from the analysis, a solution that is relatively easy to apply to small microsatellite or SNP data sets but perhaps difficult to apply to the extremely large data sets produced by genotyping‐by‐sequencing approaches. In addition, the rate of allelic dropout may vary based on the type of next‐generation method used (Flanagan & Jones, ). Possible approaches are to use a program like gbstools to estimate which SNPs in the data set are most likely suffering from allelic dropout (Cooke et al, ) or to strictly filter loci for adherence to Hardy–Weinberg equilibrium.…”

Accuracy of de novo assembly of DNA sequences from double-digest libraries varies substantially among software

“…Despite legitimate concerns about the adequacy of genome coverage by GBS-like methods for certain questions and in some systems (Lowry et al ., 2016), for many applications in population genomics GBS-like methods are likely to remain attractive for some time (McKinney, 2016). Whereas several studies have examined the consequences of laboratory and bioinformatic methods for variant identification and other downstream analyses (Shafer et al ., 2017; Flanagan & Jones, 2018; Warmuth & Ellegren, 2019), and others have suggested methods to optimize assembly parameters (Puritz et al ., 2014; Paris et al ., 2017), this investigation fills a gap in knowledge regarding the performance of de novo assembly software without the aid of additional steps.…”

Would an RRS by any other name sound as RAD?

“…But for RRS methods, many differences between techniques are minor, often implementing streamlined library preparation and cost reduction (e.g., GGRS (Chen et al., ), ezRAD (Toonen et al., )) or the use of specific restriction enzymes and adaptors designed to optimize sequencing depth, coverage, and multiplexing capacity (e.g., MSG (Andolfatto et al., ); two‐enzyme GBS (Poland et al., ); SLAF‐seq (Sun et al., ); quaddRAD (Franchini et al., )). Thus, many published methods, although prone to distinct biases or technical difficulties and subject to a myriad of downstream bioinformatic considerations (Flanagan & Jones, ; Van Dijk, Jaszczyszyn, & Thermes, ), arguably do not meet proposed criteria for publication with a unique name (e.g., NUAP, ). The recently upheld US KeyGene patent covering these methods also seems to suggest that, from a legal standpoint, they are not significantly different from one another (U.S. Patent 8,815,512 B2).…”

“…quaddRAD (Franchini et al, 2017)). Thus, many published methods, although prone to distinct biases or technical difficulties and subject to a myriad of downstream bioinformatic considerations (Flanagan & Jones, 2018;Van Dijk, Jaszczyszyn, & Thermes, 2014), arguably do not meet proposed criteria for publication with a unique name (e.g., NUAP, 2011). The recently upheld US KeyGene patent covering these methods also seems to suggest that, from a legal standpoint, they are not significantly different from one another (U.S. Patent 8,815,512 B2).…”