RefKA: A fast and efficient long-read genome assembly approach for large and complex genomes

Recent advances in long-read sequencing have the potential to produce more complete genome assemblies using sequence reads which can span repetitive regions. However, overlap based assembly methods routinely used for this data require significant computing time and resources. Here, we have developed RefKA, a reference-based approach for long read genome assembly. This approach relies on breaking up a closely related reference genome into bins, aligning k-mers unique to each bin with PacBio reads, and then assembling each bin in parallel followed by a final bin-stitching step. During benchmarking, we assembled the wheat Chinese Spring (CS) genome using publicly available PacBio reads in parallel in 168 wall hours on a 250 CPU system. The maximum RAM used was 300 Gb and the computing time was 42,000 CPU hours. The approach opens applications for the assembly of other large and complex genomes with much-reduced computing requirements. The RefKA pipeline is available at https://github.com/AppliedBioinformatics/RefKA