RAILS and Cobbler: Scaffolding and automated finishing of draft genomes using long DNA sequences

Abstract

Despite major advances in DNA sequencing technologies we do not yet have complete genome sequences. Producing high-quality, contiguous, draft assemblies de novo is of paramount importance as it informs on genetic content and organization of the genome (Pagani et al. 2012). The past decade has seen improvements in sequence throughput, a substantially lower DNA sequencing cost and increased read lengths. Whereas the base accuracy of short (currently~250 bp) read lengths such as those from Illumina have improved (>99%), the base accuracy of long sequence read platforms (Pacific Biosciences, Oxford Nanopore) remains low for generating reference-grade genome assemblies without read error correction. Gap-filling tools designed to help finish draft genomes in an automated fashion, which includes our own (Paulino et al. 2015), have been recently developed (Tsai, Otto, and Berriman 2010, Boetzer and Pirovano (2012)). They are typically designed to work with short sequencing reads, not high-quality long sequences from other draft assemblies. In many such projects that employ short sequence reads for de novo assembly, a k-mer graph assembly approach is often favored, as it effectively discards errors and spurious sequences, albeit at the cost of long-range information loss and limited ability to resolve long repeats. However, researchers routinely produce various assembly drafts varying the parameter k length in search of the most contiguous assembly. This multitude of assembly drafts is comprised of sequences with untapped potential, representing a wealth of information for gap-filling and scaffolding. Here, I make available two bioinformatics software tools, Cobbler and RAILS (Rene L Warren 2016) to exploit this information for automated finishing and scaffolding with long DNA sequences, respectively. They can be used to scaffold & finish high-quality draft genome assemblies with any long, preferably high-quality, sequences such as scaftigs/contigs from another genome draft. They both rely on accurate, long DNA sequences to patch gaps in existing genome assembly drafts. More specifically, Cobbler is a utility to automatically patch gaps (ambiguous regions in a draft assembly, represented by N's). It does so by first aligning the long sequences to the assembly, tallying the alignments and replacing N's with the sequences from these long DNA sequences. RAILS is an all-in-one scaffolder and gap-filler. Its process is similar to that of Cobbler. It scaffolds a given genome draft with the help of long DNA sequences (contig sequences are ordered/oriented using alignment information) using the scaffolding engine I originally developed for SSAKE (René L. Warren et al. 2007) and LINKS (Warren et al. 2015). The newly created gaps are automatically filled with the DNA string of the provided long DNA sequences. In a simulated long sequences experiment (1, 2.5, 5, 15 kbp sequences) designed from the human genome reference, Cobbler closed >65% of gaps in a human genome assembly draft (Table 1; test provided with the distribution, corre...

The Genome of the Beluga Whale (Delphinapterus leucas)

“…We followed a scaffolding and gap-filling methodology similar to that of the recently published bullfrog genome [ 5 ]. Gaps in our initial, supernova draft assembly were filled with Cobbler (version 0.3, Canada’s Michael Smith Genome Sciences Centre) using parameters -d 100 -i 0.95 [ 6 ] utilizing contig sequences from the ABySS assemblies generated at three kmer values ( k 95, k 100, k 105). The subsequent gap-filled assembly was initially scaffolded with RAILS (version 1.2, Canada’s Michael Smith Genome Sciences Centre) using parameters -d 100 -i 0.95 [ 6 ] using scaffold sequences from the same three ABySS assemblies.…”

“…Gaps in our initial, supernova draft assembly were filled with Cobbler (version 0.3, Canada’s Michael Smith Genome Sciences Centre) using parameters -d 100 -i 0.95 [ 6 ] utilizing contig sequences from the ABySS assemblies generated at three kmer values ( k 95, k 100, k 105). The subsequent gap-filled assembly was initially scaffolded with RAILS (version 1.2, Canada’s Michael Smith Genome Sciences Centre) using parameters -d 100 -i 0.95 [ 6 ] using scaffold sequences from the same three ABySS assemblies. Briefly, long sequences are aligned against a draft assembly using BWA-MEM (version 0.7.13), using parameters -a -t 16 [ 7 ], and the resulting alignments are parsed and inspected.…”

It’s okay to be green: Draft genome of the North American bullfrog (Rana [Lithobates] catesbeiana)

“…We assessed the improvement to the assembly after each round of scaffolding using the NG50 length metric and the number of complete and partial core eukaryotic genes (CEGs) using CEGMA, which reports a proxy metric for assembly completeness in the genic space (Parra et al 2009). Using the Synthetic Long-Reads (SLR) and the Kollector (Kucuk, et al, in press) targeted gene assembly (TGA) tool, RAILS (Warren 2016) merged over 56 thousand scaffolds; this permitted the recovery of an additional four partial CEGs, and raised the contiguity of the assembly to approximately 30 kbp (Supplemental Table S1). The most dramatic improvements to assembly contiguity and resolved CEGs were obtained using LINKS (Warren et al 2015a) and the MPET reads (NG50 increase of ~16 kbp and 10 additional complete CEGs; Supplemental Table S1), followed by the combined Kollector TGA and the lower-k whole genome assembly (~8 kbp improvement to NG50 and 9 additional complete CEGs; Supplemental Table S1).…”

“…The resulting ABySS scaffold assembly (k = 160) was re-scaffolded with RAILS version 0.1 (Warren, 2016) (ftp://ftp.bcgsc.ca/supplementary/RAILS -d 250 -i 0.99) using both Moleculo long reads and Kollector (Kucuk, et al, submitted) targeted gene reconstructions (TGA; Supplemental Table 1). In RAILS, long sequences are aligned against a draft assembly (BWA-MEM V0.7.13-r1126 (Li, 2013) -a -t16), and the 8 alignments are parsed and inspected, tracking the position and orientation of each in assembly draft sequences, satisfying minimum alignment requirements (at minimum 250 anchoring bases with 99% sequence identity or more used in this study).…”

A haplotype-resolved draft genome of the European sardine (Sardina pilchardus)

“…Further scaffolding and gap closure procedures were performed with Rails v1.2/Cobbler v0.3 pipeline script [19] to obtain the final consensus genome sequence named SP_G (ENA accession ID GCA_900499035.1) using the parameters anchoring sequence length ( −d 100) and minimum sequence identity ( −i 0.95). Three scaffolding and gap closure procedures were performed iteratively with one haplotype of the initial assembly as the assembly per se , and previous de novo assemblies from Supernova v1.2.2, (315M/100% and 450M/80% reads/barcodes).…”