Assembly of long error-prone reads using de Bruijn graphs

Abstract

The recent breakthroughs in assembling long error-prone reads were based on the overlap-layout-consensus (OLC) approach and did not utilize the strengths of the alternative de Bruijn graph approach to genome assembly. Moreover, these studies often assume that applications of the de Bruijn graph approach are limited to short and accurate reads and that the OLC approach is the only practical paradigm for assembling long error-prone reads. We show how to generalize de Bruijn graphs for assembling long error-prone reads and describe the ABruijn assembler, which combines the de Bruijn graph and the OLC approaches and results in accurate genome reconstructions.de Bruijn graph | genome assembly | single-molecule sequencing T he key challenge to the success of single-molecule sequencing (SMS) technologies lies in the development of algorithms for assembling genomes from long but inaccurate reads. The pioneer in long reads technologies, Pacific Biosciences, now produces accurate assemblies from long error-prone reads (1, 2). Goodwin et al. (3) and Loman et al. (4) demonstrated that high-quality assemblies can be obtained from even less-accurate Oxford Nanopore reads. Advances in assembly of long errorprone reads recently resulted in the accurate reconstructions of various genomes (5-10). However, as illustrated in Booher et al. (11), the problem of assembling long error-prone reads is far from being resolved even in the case of relatively small bacterial genomes.Previous studies of SMS assemblies were based on the overlaplayout-consensus (OLC) approach (12) or a similar string graph approach (13), which require an all-against-all comparison of reads (14) and remain computationally challenging (see refs. 15-17 for a discussion of the pros and cons of this approach). Moreover, there is an assumption that the de Bruijn graph approach, which has dominated genome assembly for the last decade, is inapplicable to long reads. This is a misunderstanding, because the de Bruijn graph approach, as well as its variation called the A-Bruijn graph approach, was developed to assemble rather long Sanger reads (18). There is also a misunderstanding that the de Bruijn graph approach can only assemble highly accurate reads and fails when assembling long error-prone reads. Although this is true for the original de Bruijn graph approach to assembly (15-17), the A-Bruijn graph approach was originally designed to assemble inaccurate reads as long as any similarities between reads can be reliably identified. Moreover, A-Bruijn graphs have proven to be useful even for assembling mass spectra, which represent highly inaccurate fingerprints of amino acid sequences of peptides (19,20). However, although A-Bruijn graphs have proven to be useful in assembling Sanger reads and mass spectra, the question of how to apply A-Bruijn graphs for assembling long error-prone reads remains open.de Bruijn graphs are a key algorithmic technique in genome assembly (15,(21)(22)(23)(24). In addition, de Bruijn graphs have been used for sequencing by hybridization (...

Complete genome assembly of clinical multidrug resistant Bacteroides fragilis isolates enables comprehensive identification of antimicrobial resistance genes and plasmids

“…The prospect of prediction of antimicrobial susceptibility based on resistome identification holds 58 promises for shortening time from sample to report and informing treatment decisions. Databases with 59 comprehensive reference sequences of high quality are a necessity for these purposes. Bacteroides fragilis 60…”

Yersinia canariaesp. nov., isolated from a human yersiniosis case

“…canariae NCTC 14382 T was previously sequenced by an Illumina HiSeq 2500 at Public Health England using the Nextera XP library preparation kit following a retrospective study on yersiniosis isolates cultured from patients between April 2004 and March 2018 (8 For ONT MinION data, the run metrics were inspected using NanoPlot (version 1.0) (14) before raw FAST5 files were base-called using Guppy (version 3.2.2) with the high accuracy model to FASTQ files. Adapters were trimmed from the raw reads by Porechop (version 0.2.4) using default parameters for SQK-RAD004 before the genome was de novo assembled with Flye (version 2.5) (15,16). The best assembly parameters were empirically determined to include the option flags "meta" and "plasmid" with coverage reduced to 30X for initial contig assembly based on a predicted genome size of~4.73 Mbp as informed by de novo assembly of short read Illumina data (17).…”

BELLA: Berkeley Efficient Long-Read to Long-Read Aligner and Overlapper

“…Similarly k-mers that occur more frequently than what would be expected given the sequencing depth and the error rate are likely to come from repetitive regions. It is a common practice to prune the k-mer space using various methodologies (Koren et al, 2017;Lin et al, 2016;Carvalho et al, 2016).…”

“…Following the terminology proposed by Lin et al (2016), we identify k-mers that do not exist in the genome as non-genomic, thus characterizing k-mers present in the genome as genomic. A genomic k-mer can be repeated, if it is present multiple times in the genome, or unique, if it is not.…”

“…where map(k-mer, G) is the set of locations in the genome G where k-mer can be mapped, count() function computes the cardinality of a given input set, and freq(k-mer, G, d) is the expected number of occurrences of k-mer within sequenced reads, assuming each region of G is sequenced d times (sequencing depth). In that sense, BELLA's approach to select reliable k-mers diverges sharply from how Lin et al (2016) selects their solid strings. While solid strings discards infrequent k-mers, our model discards highly-repetitive k-mers, arguing that (a) unique k-mers are sufficient to find informative overlaps, and (b) a unique k-mer has low probability of occurring frequently.…”