A pipeline for local assembly of minisatellite alleles from single-molecule sequencing data

Abstract

Motivation: The advent of Next Generation Sequencing (NGS) has led to the generation of enormous volumes of short read sequence data, cheaply and in reasonable time scales. Nevertheless, the quality of genome assemblies generated using NGS technologies has been greatly affected, compared to those generated using Sanger DNA sequencing. This is largely due to the inability of short read sequence data to scaffold repetitive structures, creating gaps, inversions and rearrangements and resulting in assemblies that are, at best, draft forms. Third generation single-molecule sequencing (SMS) technologies (e.g. Pacific Biosciences Single Molecule Real Time (SMRT) system) address this challenge by generating sequences with increased read lengths, offering the prospect to better recover these complex repetitive structures, concomitantly improving assembly quality. Results: Here, we evaluate the ability of SMS data (specifically human genome Pacific Biosciences SMRT data) to recover poorly represented repetitive sequences (specifically, GC-rich human minisatellites). To do this we designed a pipeline for the collection, processing and local assembly of single-molecule sequence data to form accurate contiguous local reconstructions. Our results show the recovery of an allele of the non-coding minisatellite MS1 (located on chromosome 1 at 1p33-35) at greater than 97% identity to reference (GRCh38) from the unprocessed sequence data of a haploid complete hydatidiform mole (CHM1) cell line. Furthermore, our assembly revealed an allele of over 500 repeat units; much larger than the reference (GRCh38), but consistent in structure with naturally occurring alleles that are segregating in human populations. This local assembly's reconstruction was validated with the release of the whole genome assemblies GCA-001297185.1 and GCA-000772585.3, where this allele occurs. Additionally, application of this pipeline to coding minisatellites in the PRDM9 and ZNF93 genes enabled recovery of high identity allele structures for these sequence regions whose length was confirmed by PCR from cell line genomic DNA. The internal repeat structure of the PRDM9 allele recovered was consistent with common human-specific alleles.