Optimizing experimental design for genome sequencing and assembly with Oxford Nanopore Technologies

Abstract

words) 6High quality reference genome sequences are the core of modern genomics. Oxford Nanopore 7Technologies (ONT) produces inexpensive DNA sequences in excess of 100,000 nucleotides but 8 error rates remain >10% and assembling these sequences, particularly for eukaryotes, is a non-9 trivial problem. To date there has been no comprehensive attempt to generate experimental 10 design for ONT genome sequencing and assembly. Here, we simulate ONT and Illumina DNA 11 sequence reads for Escherichia coli, Caenorhabditis elegans, Arabidopsis thaliana, and 12 Drosophila melanogaster. We quantify the influence of sequencing coverage, assembly software 13 and experimental design on de novo genome assembly and error correction to predict the 14 optimum sequencing strategy for these organisms. We show proof of concept using real ONT 15 data generated for the nematode Caenorhabditis remanei. ONT sequencing is inexpensive and 16 accessible, and our quantitative results will be helpful for a broad array of researchers seeking 17 guidance for de novo genome assembly projects.

Optimizing experimental design for genome sequencing and assembly with Oxford Nanopore Technologies

“…We also measured BUSCO completeness with the Diptera_odb10 for the D. melanogaster assemblies, but found that, in some instances, our assembled sequence contained a greater proportion of conserved genes than the reference sequence. We chose to focus on the Metazoan_odb10 for D. melanogaster and present both sets of statistics in the additional files in the Gigascience database [35].…”

“…The sequence read N50 indicates that 50% of the total sequenced nucleotides are in reads of that length or longer. Libraries and descriptive statistics including mean read lengths and qualities are available in the Gigascience database [35]. Using simulated libraries allowed us to study depth beyond the capability of single ONT flowcells, and to avoid possible errors created by combining libraries generated by different labs under different conditions.…”

“…Canu produced 30 contiguous pieces (LGA50 five chromosomes; Figure 10) with 98.8% of the expected Viridiplantae genes [35] identified prior to polishing [34]. Flye produced 26 contigs and 98.8% of the expected Viridiplantae genes.…”

“…Flye produced 26 contigs and 98.8% of the expected Viridiplantae genes. The Flye assembly contained many more mismatches and indels than its Canu counterpart, but this discrepancy was alleviated with Pilon polishing [35]. Following polishing with Pilon [49], the Flye-assembled sequences contained 99.1% of the expected Viridiplantae genes [34], matching that of the TAIR10 reference genome for A. thaliana.…”

“…The top-performing MaSuRCA [48] assembly produced 45 contiguous pieces, with 98.8% of the expected Viridiplantae genes identified in the sequence [34]. This was also brought up to 99.1% after polishing with Pilon [35].…”