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

Abstract

1Advances in DNA sequencing have made it feasible to gather genomic data for non-model 2 organisms and large sets of individuals, often using methods for sequencing subsets of the 3 genome. Several of these methods sequence DNA associated with endonuclease restriction 4 sites (various RAD and GBS methods). For use in taxa without a reference genome, these 5 methods rely on de novo assembly of fragments in the sequencing library. Many of the soft-6 ware options available for this application were originally developed for other assembly types 7 and we do not know their accuracy for reduced representation libraries. To address this im-8 portant knowledge gap, we simulated data from the Arabidopsis thaliana and Homo sapiens 9 genomes and compared de novo assemblies by six software programs that are commonly 10 used or promising for this purpose (ABySS, CD-HIT, Stacks, Stacks2, Velvet and VSEARCH). 11We simulated different mutation rates and types of mutations, and then applied the six 12 assemblers to the simulated datasets, varying assembly parameters. We found substantial 13 variation in software performance across simulations and parameter settings. ABySS failed 14 to recover any true genome fragments, and Velvet and VSEARCH performed poorly for most 15 simulations. Stacks and Stacks2 produced accurate assemblies of simulations containing 16 SNPs, but the addition of insertion and deletion mutations decreased their performance. 17CD-HIT was the only assembler that consistently recovered a high proportion of true genome 18 fragments. Here, we demonstrate the substantial difference in the accuracy of assemblies 19 from different software programs and the importance of comparing assemblies that result 20 from different parameter settings. 21

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

“…CD‐HIT was developed for assembling protein sequences, but was later extended for nucleotide sequences, and it is used in the analysis pipeline dDocent (Puritz et al, ). We used dDocent's data reduction step that retains only one copy of each unique sequence for assembly to reduce computational time (this script can be found in the Dryad repository: https://doi.org/10.5061/dryad.8tr03f8, LaCava et al, ).…”

“…We used indsdemultiplexing.py to demultiplex the simulated reads and readstrim.py to trim the barcodes from the reads, resulting in reads of 94 bp that all began with the EcoRI restriction site. Our wrapper functions for each simulation, modified from run_ddradseq_chain.sh in ddRADseqTools, are included in the Dryad repository: https://doi.org/10.5061/dryad.8tr03f8, LaCava et al ().…”

“…We constructed GBS assemblies using each assembler for the nine simulated data sets for A. thaliana and H. sapiens . We used a custom Perl script to compare the assembled contigs with the known fragments from the in‐silico digestion of genomes to determine assembly accuracy using two metrics (this script can be found in the Dryad repository: https://doi.org/10.5061/dryad.8tr03f8, LaCava et al, ). To evaluate how completely each assembler recovered the original genome fragments (loci), we counted the number of true genome fragments that were represented in the assembly (completeness criterion).…”

“…Simulated reads, assembler outputs, and scripts for simulations, assembly, and analysis are available from the Dryad Digital Repository: https://doi.org/10.5061/dryad.8tr03f8 (LaCava et al, ).…”

Stacking up RADSeq assembly programs: From complete hit to completely abysmal

“…Despite many studies using RRL methods, the field lacks guidelines for best practices, and researchers have experimented with many different de novo assembly pipelines. A survey of 100 studies with de novo assembly of double‐digest RRLs published between February 2015 and October 2017 used 24 unique assembly pipelines (LaCava et al, ). That is approximately one new assembler per four similarly designed studies!…”