ddrage: A data set generator to evaluate ddRADseq analysis software

Abstract

High-throughput sequencing makes it possible to evaluate thousands of genetic markers across genomes and populations. Reduced-representation sequencing approaches, like double-digest restriction site-associated DNA sequencing (ddRADseq), are frequently applied to screen for genetic variation. In particular in nonmodel organisms where whole-genome sequencing is not yet feasible, ddRADseq has become popular as it allows genomewide assessment of variation patterns even in the absence of other genomic resources. However, while many tools are available for the analysis of ddRADseq data, few options exist to simulate ddRADseq data in order to evaluate the accuracy of downstream tools. The available tools either focus on the optimization of ddRAD experiment design or do not provide the information necessary for a detailed evaluation of different ddRAD analysis tools. For this task, a ground truth, that is, the underlying information of all effects in the data set, is required. Therefore, we here present ddrage, the ddRAD Data Set Generator, that allows both developers and users to evaluate their ddRAD analysis software. ddrage allows the user to adjust many parameters such as coverage and rates of mutations, sequencing errors or allelic dropouts, in order to generate a realistic simulated ddRADseq data set for given experimental scenarios and organisms. The simulated reads can be easily processed with available analysis software such as stacks or pyrad and evaluated against the underlying parameters used to generate the data to gauge the impact of different parameter values used during downstream data processing.

Population genomics analysis with RAD, reprised: Stacks 2

“…There are several software packages designed around this initial estimation of RAD loci, including SIMRLLS (Eaton et al 2016), SIMRAD (Lepais and Weir 2014), DDRADSEQTOOLS (Mora-Márquez et al 2017), DDRAGE (Timm et al 2018), PREDRAD (Herrera et al 2015), andRADINITIO (Rivera-Colón et al 2021). All these programs base their estimation on counting restriction enzyme cut sites on a reference sequence but vary in the way they account for the variation introduced by the presence of repetitive loci, natural polymorphism, and technical error.…”

Optimizing ddRADseq in Non-Model Species: A Case Study in Eucalyptus dunnii Maiden

“…These approaches involve digesting the DNA with restriction enzymes and then sequencing a specific size-selected range of generated fragments. Aiming to ameliorate some of the weaknesses of the original RADseq, specifically with regard to the dependence of the length of the generated fragments on random shearing effects [5,6], researchers have developed many derived methodologies, including 2b-RADseq [7], ezRADseq [8] and ddRADseq [9]. Double-digest restriction site-associated DNA sequencing, or ddRADseq, uses two different restriction enzymes to cut the DNA: one rare cutter (i.e., an enzyme with a large recognition site) and a frequent cutter.…”

“…Double-digest restriction site-associated DNA sequencing, or ddRADseq, uses two different restriction enzymes to cut the DNA: one rare cutter (i.e., an enzyme with a large recognition site) and a frequent cutter. Only the fragments falling between both restriction sites and within a specific size range are sequenced [6]. This reduces the depth of sequencing needed to reach optimal coverage, as well as the percentage of missing data, in comparison with RADseq.…”

Simulation with RADinitio improves RADseq experimental design and sheds light on sources of missing data

“…While it also simulates variants for multiple samples, these are independent across individuals, and have only fixed effects over allele dropout. Finally, ddrage (Timm, Weigand, Weiss, Leese, & Rahmann, 2018) simulates ddRADseq data under multiple library configurations, in both a de novo and reference context. It can also generate variants that are shared across individuals.…”