Pre-Assembly NGS Correction of ONT Reads Achieves HiFi-Level Assembly Quality

Abstract

BackgroundRecently developed hybrid assemblies can achieve Telomere-to-Telomere (T2T) completeness of some chromosomes. However, such approaches involve sequencing a large volume of both Pacific Biosciences high-fidelity (HiFi) and Oxford Nanopore Technologies (ONT) sequencing reads. Along with this, third-generation sequencing techniques are rapidly advancing, increasing the available length and accuracy. To reduce the final cost of genome assembly, here we investigated the possibility of assembly from low-coverage samples and with only ONT corrected by Next-Generation Sequencing (NGS) sequencing reads. We demonstrated that ONT-based assembly approaches corrected by NGS can achieve performance metrics comparable to more expensive hybrid approaches based on HiFi sequencing. ResultsWe investigated the assembly of different chromosomes of HG002 and the low-coverage performance of state-of-the-art hybrid assembly tools, including Verkko and Hifiasm, and ONT-based assemblers Shasta and Flye. Our study shows that even with the best NG50, Hifiasm has many misassemblies in the centromere region, raising doubts about its T2T capabilities. Additionally, ONT-based assemblies perform well at low coverage, offering a cost-effective solution. To achieve better assembly quality, we evaluated the performance of different NGS technologies across various aspects of hybrid genome assembly, including pre-assembly correction, haplotype phasing, and polishing, and found them to be similarly effective. For the NGS technology stLFR, we proposed two-round assembly methods that utilize stLFR linked-read data to achieve assembly phasing performance comparable to that obtained with trio data. ConclusionWe found that assemblies based on ONT corrected by NGS can match or exceed HiFi-based assemblies in accuracy. It appears that no assembly tool investigated here can achieve T2T completeness, even with 50X coverage. Therefore, we recommend using a combination of regular R9 or R10 simplex ONT reads and accurate NGS reads for assembly without aiming for T2T completeness. These ONT-based assemblies perform well even at low coverage, offering a cost-effective solution.