GapPredict – A Language Model for Resolving Gaps in Draft Genome Assemblies

Abstract

Short-read DNA sequencing instruments can yield over 10 12 bases per run, typically composed of reads 150 bases long. Despite this high throughput, de novo assembly algorithms have difficulty reconstructing contiguous genome sequences using short reads due to both repetitive and difficult-to-sequence regions in these genomes. Some of the short read assembly challenges are mitigated by scaffolding assembled sequences using paired-end reads. However, unresolved sequences in these scaffolds appear as "gaps". Here, we introduce GapPredict -an implementation of a proof of concept that uses a character-level language model to predict unresolved nucleotides in scaffold gaps. We benchmarked GapPredict against the state-of-the-art gap-filling tool Sealer, and observed that the former can fill 65.6% of the sampled gaps that were left unfilled by the latter with high similarity to the reference genome, demonstrating the practical utility of deep learning approaches to the gap-filling problem in genome assembly.

Complete mitochondrial genome of Gnathostoma binucleatum

“…Subsequently, Geneious performed autoannotation using the NC_034239.1 reference sequence as a gene transfer guide. To refine the draft genome, GapPredict’s machine learning algorithm filled gaps and resolved ambiguous bases, which were then confirmed manually via raw read alignment and genomic context ( 5 ). Protein-coded gene calls were confirmed to have appropriate open reading frames by using Expasy Translate with the invertebrate mitochondrial code and then queried in the non-redundant GenBank database.…”