PMERGE: Computational filtering of paralogous sequences from RAD‐seq data

Restriction‐site associated DNA sequencing (RAD‐seq) can identify and score thousands of genetic markers from a group of samples for population‐genetics studies. One challenge of de novo RAD‐seq analysis is to distinguish paralogous sequence variants (PSVs) from true single‐nucleotide polymorphisms (SNPs) associated with orthologous loci. In the absence of a reference genome, it is difficult to differentiate true SNPs from PSVs, and their impact on downstream analysis remains unclear. Here, we introduce a network‐based approach, PMERGE that connects fragments based on their DNA sequence similarity to identify probable PSVs. Applying our method to de novo RAD‐seq data from 150 Atlantic salmon (Salmo salar) samples collected from 15 locations across the Southern Newfoundland coast allowed the identification of 87% of total PSVs identified through alignment to the Atlantic salmon genome. Removal of these paralogs altered the inferred population structure, highlighting the potential impact of filtering in RAD‐seq analysis. PMERGE is also applied to a green crab (Carcinus maenas) data set consisting of 242 samples from 11 different locations and was successfully able to identify and remove the majority of paralogous loci (62%). The PMERGE software can be run as part of the widely used Stacks analysis package.     

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

Advances in DNA sequencing have made it feasible to gather genomic data for non‐model organisms and large sets of individuals, often using methods for sequencing subsets of the genome. Several of these methods sequence DNA associated with endonuclease restriction sites (various RAD and GBS methods). For use in taxa without a reference genome, these methods rely on de novo assembly of fragments in the sequencing library. Many of the software options available for this application were originally developed for other assembly types and we do not know their accuracy for reduced representation libraries. To address this important knowledge gap, we simulated data from the Arabidopsis thaliana and Homo sapiens genomes and compared de novo assemblies by six software programs that are commonly used or promising for this purpose (ABySS , CD‐HIT , Stacks , Stacks2 , Velvet and VSEARCH ). We simulated different mutation rates and types of mutations, and then applied the six assemblers to the simulated data sets, varying assembly parameters. We found substantial variation in software performance across simulations and parameter settings. ABySS failed to recover any true genome fragments, and Velvet and VSEARCH performed poorly for most simulations. Stacks and Stacks2 produced accurate assemblies of simulations containing SNPs, but the addition of insertion and deletion mutations decreased their performance. CD‐HIT was the only assembler that consistently recovered a high proportion of true genome fragments. Here, we demonstrate the substantial difference in the accuracy of assemblies from different software programs and the importance of comparing assemblies that result from different parameter settings.     

Investigating the genomic basis of discrete phenotypes using a Pool‐Seq‐only approach: New insights into the genetics underlying colour variation in diverse taxa

While large‐scale genomic approaches are increasingly revealing the genetic basis of polymorphic phenotypes such as colour morphs, such approaches are almost exclusively conducted in species with high‐quality genomes and annotations. Here, we use Pool‐Seq data for both genome assembly and SNP frequency estimation, followed by scanning for F_ST outliers to identify divergent genomic regions. Using paired‐end, short‐read sequencing data from two groups of individuals expressing divergent phenotypes, we generate a de novo rough‐draft genome, identify SNPs and calculate genomewide F_ST differences between phenotypic groups. As genomes generated by Pool‐Seq data are highly fragmented, we also present an approach for super‐scaffolding contigs using existing protein‐coding data sets. Using this approach, we reanalysed genomic data from two recent studies of birds and butterflies investigating colour pattern variation and replicated their core findings, demonstrating the accuracy and power of a Pool‐Seq‐only approach. Additionally, we discovered new regions of high divergence and new annotations that together suggest novel parallels between birds and butterflies in the origins of their colour pattern variation.     

Development of a universal double‐digest RAD sequencing approach for a group of nonmodel,ecologically and economically important insect and fish taxa

The generation of genome‐scale data is critical for a wide range of questions in basic biology using model organisms, but also in questions of applied biology in nonmodel organisms (agriculture, natural resources, conservation and public health biology). Using a genome‐scale approach on a diverse group of nonmodel organisms and with the goal of lowering costs of the method, we modified a multiplexed, high‐throughput genomic scan technique utilizing two restriction enzymes. We analysed several pairs of restriction enzymes and completed double‐digestion RAD sequencing libraries for nine different species and five genera of insects and fish. We found one particular enzyme pair produced consistently higher number of sequence‐able fragments across all nine species. Building libraries off this enzyme pair, we found a range of usable SNPs between 4000 and 37 000 SNPS per species and we found a greater number of usable SNPs using reference genomes than de novo pipelines in STACKS. We also found fewer reads in the Read 2 fragments from the paired‐end Illumina Hiseq run. Overall, the results of this study provide empirical evidence of the utility of this method for producing consistent data for diverse nonmodel species and suggest specific considerations for sequencing analysis strategies.     

Mass production of SNP markers in a nonmodel passerine bird through RAD sequencing and contig mapping to the zebra finch genome

Here, we present an adaptation of restriction‐site‐associated DNA sequencing (RAD‐seq) to the Illumina HiSeq2000 technology that we used to produce SNP markers in very large quantities at low cost per unit in the Réunion grey white‐eye (Zosterops borbonicus), a nonmodel passerine bird species with no reference genome. We sequenced a set of six pools of 18–25 individuals using a single sequencing lane. This allowed us to build around 600 000 contigs, among which at least 386 000 could be mapped to the zebra finch (Taeniopygia guttata) genome. This yielded more than 80 000 SNPs that could be mapped unambiguously and are evenly distributed across the genome. Thus, our approach provides a good illustration of the high potential of paired‐end RAD sequencing of pooled DNA samples combined with comparative assembly to the zebra finch genome to build large contigs and characterize vast numbers of informative SNPs in nonmodel passerine bird species in a very efficient and cost‐effective way.     

RADProc: A computationally efficient de novo locus assembler for population studies using RADseq data

Restriction site‐associated DNA sequencing (RADseq) is a powerful tool for genotyping of individuals, but the identification of loci and assignment of sequence reads is a crucial and often challenging step. The optimal parameter settings for a given de novo RADseq assembly vary between data sets and can be difficult and computationally expensive to determine. Here, we introduce RADProc, a software package that uses a graph data structure to represent all sequence reads and their similarity relationships. Storing sequence–comparison results in a graph eliminates unnecessary and redundant sequence similarity calculations. De novo locus formation for a given parameter set can be performed on the precomputed graph, making parameter sweeps far more efficient. RADProc also uses a clustering approach for faster nucleotide‐distance calculation. The performance of RADProc compares favourably with that of the widely used Stacks software. The run‐time comparisons between RADProc and Stacks for 32 different parameter settings using 20 green‐crab (Carcinus maenas) samples showed that RADProc took as little as 2 hr 40 min compared to 78 hr by Stacks, while 16 brown trout (Salmo trutta L.) samples were processed by RADProc and Stacks in 23 and 263 hr, respectively. Comparisons of the de novo loci formed, and catalog built using both the methods demonstrate that the improvement in processing speeds achieved by RADProc does not affect much the actual loci formed and the results of downstream analyses based on those loci.     

SNP discovery using Paired‐End RAD‐tag sequencing on pooled genomic DNA of Sisymbrium austriacum (Brassicaceae)

Single nucleotide polymorphisms SNPs are rapidly replacing anonymous markers in population genomic studies, but their use in non model organisms is hampered by the scarcity of cost‐effective approaches to uncover genome‐wide variation in a comprehensive subset of individuals. The screening of one or only a few individuals induces ascertainment bias. To discover SNPs for a population genomic study of the Pyrenean rocket (Sisymbrium austriacum subsp. chrysanthum), we undertook a pooled RAD‐PE (Restriction site Associated DNA Paired‐End sequencing) approach. RAD tags were generated from the PstI‐digested pooled genomic DNA of 12 individuals sampled across the species distribution range and paired‐end sequenced using Illumina technology to produce ~24.5 Mb of sequences, covering ~7% of the specie's genome. Sequences were assembled into ~76 000 contigs with a mean length of 323 bp (N₅₀ = 357 bp, sequencing depth = 24x). In all, >15 000 SNPs were called, of which 47% were annotated in putative genic regions based on homology with the Arabidopsis thaliana genome. Gene ontology (GO) slim categorization demonstrated that the identified SNPs covered extant genic variation well. The validation of 300 SNPs on a larger set of individuals using a KASPar assay underpinned the utility of pooled RAD‐PE as an inexpensive genome‐wide SNP discovery technique (success rate: 87%). In addition to SNPs, we discovered >600 putative SSR markers.     

Microhaplotypes provide increased power from short‐read DNA sequences for relationship inference

The accelerating rate at which DNA sequence data are now generated by high‐throughput sequencing instruments provides both opportunities and challenges for population genetic and ecological investigations of animals and plants. We show here how the common practice of calling genotypes from a single SNP per sequenced region ignores substantial additional information in the phased short‐read sequences that are provided by these sequencing instruments. We target sequenced regions with multiple SNPs in kelp rockfish (Sebastes atrovirens) to determine “microhaplotypes” and then call these microhaplotypes as alleles at each locus. We then demonstrate how these multi‐allelic marker data from such loci dramatically increase power for relationship inference. The microhaplotype approach decreases false‐positive rates by several orders of magnitude, relative to calling bi‐allelic SNPs, for two challenging analytical procedures, full‐sibling and single parent–offspring pair identification. We also show how the identification of half‐sibling pairs requires so much data that physical linkage becomes a consideration, and that most published studies that attempt to do so are dramatically underpowered. The advent of phased short‐read DNA sequence data, in conjunction with emerging analytical tools for their analysis, promises to improve efficiency by reducing the number of loci necessary for a particular level of statistical confidence, thereby lowering the cost of data collection and reducing the degree of physical linkage amongst markers used for relationship estimation. Such advances will facilitate collaborative research and management for migratory and other widespread species.     

Finding the right coverage: the impact of coverage and sequence quality on single nucleotide polymorphism genotyping error rates

Restriction‐enzyme‐based sequencing methods enable the genotyping of thousands of single nucleotide polymorphism (SNP) loci in nonmodel organisms. However, in contrast to traditional genetic markers, genotyping error rates in SNPs derived from restriction‐enzyme‐based methods remain largely unknown. Here, we estimated genotyping error rates in SNPs genotyped with double digest RAD sequencing from Mendelian incompatibilities in known mother–offspring dyads of Hoffman's two‐toed sloth (Choloepus hoffmanni) across a range of coverage and sequence quality criteria, for both reference‐aligned and de novo‐assembled data sets. Genotyping error rates were more sensitive to coverage than sequence quality and low coverage yielded high error rates, particularly in de novo‐assembled data sets. For example, coverage ≥5 yielded median genotyping error rates of ≥0.03 and ≥0.11 in reference‐aligned and de novo‐assembled data sets, respectively. Genotyping error rates declined to ≤0.01 in reference‐aligned data sets with a coverage ≥30, but remained ≥0.04 in the de novo‐assembled data sets. We observed approximately 10‐ and 13‐fold declines in the number of loci sampled in the reference‐aligned and de novo‐assembled data sets when coverage was increased from ≥5 to ≥30 at quality score ≥30, respectively. Finally, we assessed the effects of genotyping coverage on a common population genetic application, parentage assignments, and showed that the proportion of incorrectly assigned maternities was relatively high at low coverage. Overall, our results suggest that the trade‐off between sample size and genotyping error rates be considered prior to building sequencing libraries, reporting genotyping error rates become standard practice, and that effects of genotyping errors on inference be evaluated in restriction‐enzyme‐based SNP studies.     

A resource of genome‐wide single‐nucleotide polymorphisms generated by RAD tag sequencing in the critically endangered European eel

Reduced representation genome sequencing such as restriction‐site‐associated DNA (RAD) sequencing is finding increased use to identify and genotype large numbers of single‐nucleotide polymorphisms (SNPs) in model and nonmodel species. We generated a unique resource of novel SNP markers for the European eel using the RAD sequencing approach that was simultaneously identified and scored in a genome‐wide scan of 30 individuals. Whereas genomic resources are increasingly becoming available for this species, including the recent release of a draft genome, no genome‐wide set of SNP markers was available until now. The generated SNPs were widely distributed across the eel genome, aligning to 4779 different contigs and 19 703 different scaffolds. Significant variation was identified, with an average nucleotide diversity of 0.00529 across individuals. Results varied widely across the genome, ranging from 0.00048 to 0.00737 per locus. Based on the average nucleotide diversity across all loci, long‐term effective population size was estimated to range between 132 000 and 1 320 000, which is much higher than previous estimates based on microsatellite loci. The generated SNP resource consisting of 82 425 loci and 376 918 associated SNPs provides a valuable tool for future population genetics and genomics studies and allows for targeting specific genes and particularly interesting regions of the eel genome.     

GIbPSs: a toolkit for fast and accurate analyses of genotyping‐by‐sequencing data without a reference genome

Genotyping‐by‐sequencing (GBS) and related methods are increasingly used for studies of non‐model organisms from population genetic to phylogenetic scales. We present GIbPSs, a new genotyping toolkit for the analysis of data from various protocols such as RAD, double‐digest RAD, GBS, and two‐enzyme GBS without a reference genome. GIbPSs can handle paired‐end GBS data and is able to assign reads from both strands of a restriction fragment to the same locus. GIbPSs is most suitable for population genetic and phylogeographic analyses. It avoids genotyping errors due to indel variation by identifying and discarding affected loci. GIbPSs creates a genotype database that offers rich functionality for data filtering and export in numerous formats. We performed comparative analyses of simulated and real GBS data with GIbPSs and another program, pyRAD. This program accounts for indel variation by aligning homologous sequences. GIbPSs performed better than pyRAD in several aspects. It required much less computation time and displayed higher genotyping accuracy. GIbPSs retained smaller numbers of loci overall in analyses of real GBS data. It nevertheless delivered more complete genotype matrices with greater locus overlap between individuals and greater numbers of loci sampled in all individuals.     

Estimation of population allele frequencies from next‐generation sequencing data: pool‐versus individual‐based genotyping

Molecular markers produced by next‐generation sequencing (NGS) technologies are revolutionizing genetic research. However, the costs of analysing large numbers of individual genomes remain prohibitive for most population genetics studies. Here, we present results based on mathematical derivations showing that, under many realistic experimental designs, NGS of DNA pools from diploid individuals allows to estimate the allele frequencies at single nucleotide polymorphisms (SNPs) with at least the same accuracy as individual‐based analyses, for considerably lower library construction and sequencing efforts. These findings remain true when taking into account the possibility of substantially unequal contributions of each individual to the final pool of sequence reads. We propose the intuitive notion of effective pool size to account for unequal pooling and derive a Bayesian hierarchical model to estimate this parameter directly from the data. We provide a user‐friendly application assessing the accuracy of allele frequency estimation from both pool‐ and individual‐based NGS population data under various sampling, sequencing depth and experimental error designs. We illustrate our findings with theoretical examples and real data sets corresponding to SNP loci obtained using restriction site–associated DNA (RAD) sequencing in pool‐ and individual‐based experiments carried out on the same population of the pine processionary moth (Thaumetopoea pityocampa). NGS of DNA pools might not be optimal for all types of studies but provides a cost‐effective approach for estimating allele frequencies for very large numbers of SNPs. It thus allows comparison of genome‐wide patterns of genetic variation for large numbers of individuals in multiple populations.     

multi‐dice: r package for comparative population genomic inference under hierarchical co‐demographic models of independent single‐population size changes

Population genetic data from multiple taxa can address comparative phylogeographic questions about community‐scale response to environmental shifts, and a useful strategy to this end is to employ hierarchical co‐demographic models that directly test multi‐taxa hypotheses within a single, unified analysis. This approach has been applied to classical phylogeographic data sets such as mitochondrial barcodes as well as reduced‐genome polymorphism data sets that can yield 10,000s of SNPs, produced by emergent technologies such as RAD‐seq and GBS. A strategy for the latter had been accomplished by adapting the site frequency spectrum to a novel summarization of population genomic data across multiple taxa called the aggregate site frequency spectrum (aSFS), which potentially can be deployed under various inferential frameworks including approximate Bayesian computation, random forest and composite likelihood optimization. Here, we introduce the r package multi‐dice , a wrapper program that exploits existing simulation software for flexible execution of hierarchical model‐based inference using the aSFS, which is derived from reduced genome data, as well as mitochondrial data. We validate several novel software features such as applying alternative inferential frameworks, enforcing a minimal threshold of time surrounding co‐demographic pulses and specifying flexible hyperprior distributions. In sum, multi‐dice provides comparative analysis within the familiar R environment while allowing a high degree of user customization, and will thus serve as a tool for comparative phylogeography and population genomics.     

Identification of SNP markers for inferring phylogeny in temperate bamboos (Poaceae: Bambusoideae) using RAD sequencing

Phylogenetic relationships among temperate species of bamboo are difficult to resolve, owing to both the challenge of detecting sufficiently variable markers and their polyploid history. Here, we use restriction site–associated DNA sequencing to identify candidate loci with fixed allelic differences segregating between and within two temperate species of bamboos: Arundinaria faberi and Yushania brevipaniculata. Approximately 27 million paired‐end sequencing reads were generated across four samples. From pooled data, we assembled 67 685 and 70 668 de novo contigs from partial overlap among paired‐end reads, with an average length of 240 and 241 bp for the two species, respectively, which were used to investigate functional classification of RAD tags in a blastx search. Analysed separately by population, we recovered 29 443 putatively orthologous RAD tags shared across the four sampled populations, containing 28 023 sequence variants, of which c. 13 000 are segregating between species, and c. 3000 segregating between populations within each species. Analyses based on these RAD tags yielded robust phylogenetic inferences, even with data set constructed from surprisingly few loci. This study illustrates the potential for reduced‐representation genome data to resolve difficult phylogenetic relationships in temperate bamboos.     

Identifying the causes and consequences of assembly gaps using a multiplatform genome assembly of a bird‐of‐paradise

Genome assemblies are currently being produced at an impressive rate by consortia and individual laboratories. The low costs and increasing efficiency of sequencing technologies now enable assembling genomes at unprecedented quality and contiguity. However, the difficulty in assembling repeat‐rich and GC‐rich regions (genomic “dark matter”) limits insights into the evolution of genome structure and regulatory networks. Here, we compare the efficiency of currently available sequencing technologies (short/linked/long reads and proximity ligation maps) and combinations thereof in assembling genomic dark matter. By adopting different de novo assembly strategies, we compare individual draft assemblies to a curated multiplatform reference assembly and identify the genomic features that cause gaps within each assembly. We show that a multiplatform assembly implementing long‐read, linked‐read and proximity sequencing technologies performs best at recovering transposable elements, multicopy MHC genes, GC‐rich microchromosomes and the repeat‐rich W chromosome. Telomere‐to‐telomere assemblies are not a reality yet for most organisms, but by leveraging technology choice it is now possible to minimize genome assembly gaps for downstream analysis. We provide a roadmap to tailor sequencing projects for optimized completeness of both the coding and noncoding parts of nonmodel genomes.     

Population genomic analyses from low‐coverage RAD‐Seq data: a case study on the non‐model cucurbit bottle gourd

Restriction site‐associated DNA sequencing (RAD‐Seq), a next‐generation sequencing‐based genome ‘complexity reduction’ protocol, has been useful in population genomics in species with a reference genome. However, the application of this protocol to natural populations of genomically underinvestigated species, particularly under low‐to‐medium sequencing depth, has not been well justified. In this study, a Bayesian method was developed for calling genotypes from an F₂ population of bottle gourd [Lagenaria siceraria (Mol.) Standl.] to construct a high‐density genetic map. Low‐depth genome shotgun sequencing allowed the assembly of scaffolds/contigs comprising approximately 50% of the estimated genome, of which 922 were anchored for identifying syntenic regions between species. RAD‐Seq genotyping of a natural population comprising 80 accessions identified 3226 single nuclear polymorphisms (SNPs), based on which two sub‐gene pools were suggested for association with fruit shape. The two sub‐gene pools were moderately differentiated, as reflected by the Hudson's F_ST value of 0.14, and they represent regions on LG7 with strikingly elevated F_ST values. Seven‐fold reduction in heterozygosity and two times increase in LD (r²) were observed in the same region for the round‐fruited sub‐gene pool. Outlier test suggested the locus LX3405 on LG7 to be a candidate site under selection. Comparative genomic analysis revealed that the cucumber genome region syntenic to the high F_ST island on LG7 harbors an ortholog of the tomato fruit shape gene OVATE. Our results point to a bright future of applying RAD‐Seq to population genomic studies for non‐model species even under low‐to‐medium sequencing efforts. The genomic resources provide valuable information for cucurbit genome research.     

Population structure of Atlantic mackerel inferred from RAD‐seq‐derived SNP markers: effects of sequence clustering parameters and hierarchical SNP selection

Restriction‐site‐associated DNA sequencing (RAD‐seq) and related methods are revolutionizing the field of population genomics in nonmodel organisms as they allow generating an unprecedented number of single nucleotide polymorphisms (SNPs) even when no genomic information is available. Yet, RAD‐seq data analyses rely on assumptions on nature and number of nucleotide variants present in a single locus, the choice of which may lead to an under‐ or overestimated number of SNPs and/or to incorrectly called genotypes. Using the Atlantic mackerel (Scomber scombrus L.) and a close relative, the Atlantic chub mackerel (Scomber colias), as case study, here we explore the sensitivity of population structure inferences to two crucial aspects in RAD‐seq data analysis: the maximum number of mismatches allowed to merge reads into a locus and the relatedness of the individuals used for genotype calling and SNP selection. Our study resolves the population structure of the Atlantic mackerel, but, most importantly, provides insights into the effects of alternative RAD‐seq data analysis strategies on population structure inferences that are directly applicable to other species.