Performance and precision of double digestion RAD (ddRAD) genotyping in large multiplexed datasets of marine fish species

Abstract

The development of Genotyping-By-Sequencing (GBS) technologies enables cost-effective analysis of large numbers of Single Nucleotide Polymorphisms (SNPs), especially in "non-model" species. Nevertheless, as such technologies enter a mature phase, biases and errors inherent to GBS are becoming evident. Here, we evaluated the performance of double digest Restriction enzyme Associated DNA (ddRAD) sequencing in SNP genotyping studies including high number of samples. Datasets of sequence data were generated from three marine teleost species (>5500 samples, >2.5 × 10 bases in total), using a standardized protocol. A common bioinformatics pipeline based on STACKS was established, with and without the use of a reference genome. We performed analyses throughout the production and analysis of ddRAD data in order to explore (i) the loss of information due to heterogeneous raw read number across samples; (ii) the discrepancy between expected and observed tag length and coverage; (iii) the performances of reference based vs. de novo approaches; (iv) the sources of potential genotyping errors of the library preparation/bioinformatics protocol, by comparing technical replicates. Our results showed use of a reference genome and a posteriori genotype correction improved genotyping precision. Individual read coverage was a key variable for reproducibility; variance in sequencing depth between loci in the same individual was also identified as an important factor and found to correlate to tag length. A comparison of downstream analysis carried out with ddRAD vs single SNP allele specific assay genotypes provided information about the levels of genotyping imprecision that can have a significant impact on allele frequency estimations and population assignment. The results and insights presented here will help to select and improve approaches to the analysis of large datasets based on RAD-like methodologies.

Genome wide analysis reveals genetic divergence between Goldsinny wrasse populations

“…S8a-c). This is in line with the earlier observation that variance in read coverage between individuals and between loci in the same individual introduce biases (Maroso et al 2018). However, based on results from this study, larger sample sizes are likely also needed in search of reliable outliers between populations (Fig.…”

“…On the other hand, this in turn would have come on the expense of genomic coverage, and thus possibility to detect (any) adaptive outliers (Hoban et al 2016), and could also introduce other type of errors and biases (see e.g. Maroso et al 2018;Wright et al 2019;Graham et al 2020).…”

Characterizing the genetic structure of introduced Nile tilapia (Oreochromis niloticus) strains in Tanzania using double digest RAD sequencing

“…Both RAD and ddRAD-seq have been widely applied in aquaculture breeding and genetics studies (Robledo et al 2018). In particular, ddRAD-seq has been applied for genotyping large multiplexed datasets (e.g., Maroso et al 2018), construction of genetic linkage maps (e.g., Recknagel et al 2013;Oral et al 2017), analyzing life history traits (e.g., Pukk 2016), mapping sex determining loci (e.g., Palaiokostas et al 2015;Brown et al 2016), genomic predictions and genome-wide association studies (e.g., Barria et al 2018), assessing genetic diversity (e.g., Hosoya et al 2018;Tony et al 2015;Siccha-Ramirez et al 2018;Torati et al 2019), phylogeography (e.g., Stobie et al 2018, and species identification in tilapias (Syaifudin et al 2019).…”

Utilizing field collected insects for next generation sequencing: Effects of sampling, storage, and DNA extraction methods

“…We assessed three main measures as proxies for evaluating ddRAD sequencing success: number of polymorphic loci, sequencing read depth, and levels of missing loci between treatments within a species. While number of loci is often utilized as a standard measurement of sequencing success (Graham et al, ), depth of coverage (number of sequence reads for a given locus) is also an important metric of sequence quality, as higher depth allows for greater detection of sequencing errors, heterozygous loci, and differences between individuals and populations (Maroso et al, ; Sims, Sudbery, Ilott, Heger, & Ponting, ). Quantifying the amount of missing data is also an important aspect of assessing RAD success, as there is a finite number of sequencing reads spread across multiple individuals during sequencing, and the assembly of genomes, either de novo or mapped to a reference, depends on sequence similarity (low levels of missing data) between specimens (Catchen, Amores, & Hohenlohe, ).…”

“…We assessed three main measures as proxies for evaluating ddRAD sequencing success: number of polymorphic loci, sequencing read depth, and levels of missing loci between treatments within a species. While number of loci is often utilized as a standard measurement of sequencing success (Graham et al, 2015), depth of coverage (number of sequence reads for a given locus) is also an important metric of sequence quality, as higher depth allows for greater detection of sequencing errors, heterozygous loci, and differences between individuals and populations (Maroso et al, 2018;Sims, Sudbery, Ilott, Heger, & Ponting, 2014).…”