NanoPack2: population-scale evaluation of long-read sequencing data

1Long-read sequencing from Pacific Biosciences and Oxford Nanopore Technologies (ONT) has evolved from single genomes and small groups of individuals to large population-scale cohorts. Simultaneously, the increasing economic cost and climate impact of computational tasks also necessitate more efficient bioinformatic methods for data quality assessment and processing. However, several tools have been developed for the quality assessment of long-read sequencing data without scaling to populations of >100 genomes. This article presents newly developed tools that fulfill this need and efficiently assess characteristics relevant to long-read genome sequencing, including alignments spanning structural variants and phasing read alignments. Phasing, i.e. assigning each sequenced fragment to a parental haplotype by identifying co-occurring variants, is critical in identifying potential functional variants in association studies and for the pathogenicity of putative compound heterozygous variation. Furthermore, we present an update on NanoPlot and NanoComp from the NanoPack tools.

2Improvements to NanoPlot and NanoComp are, among code optimizations, the generation of additional plots, using dynamic HTML plots from the Plotly library, and enabling further exploration by the end users (Supplementary Fig. S1). The tools now also support input using the programming language agnostic Arrow data format. A binary is provided to efficiently generate Arrow files from BAM/CRAM alignments. Chopper is a tool that combines the utility of NanoFilt and NanoLyse, for filtering sequencing reads based on quality, length, and contaminating sequences, delivers a 7-fold speed up compared to the Python implementation, making use of the Rust-Bio library and Rust bindings to minimap2.

3Summarization of long-read sequencing experiments using NanoStat is too slow considering the yields that are nowadays common with nanopore sequencing. Cramino, using rust-htslib, provides a much faster alternative for gathering metrics based on the data output, mean coverage, the number of reads, their mean and median length, and sequence identity relative to the reference genome. Long reads span structural variants, and penalizing the read accuracy for a large gap is undesirable. For this reason, Cramino calculates the gap-compressed identity, defined as the edit distance relative to the read length, while counting consecutive alignment gaps as just one difference (Supplementary Methods). Cramino allows filtering on read length and optionally outputs a rudimentary evaluation of the karyotype and biological sex by calculating normalized read counts per chromosome, calculates the MD5 checksum to control for data integrity, provides metrics of the read phasing performance, and can provide metrics on the number of spliced exons for long-read transcriptomics. Importantly, Cramino remains compatible with the rich visualizations from NanoPlot and NanoComp by generating output in the Arrow format on top of the optional lightweight histograms for read length and read identity from Cramino itself. For ONT human genome sequencing with 50× coverage using four cores for BAM/CRAM decompression, Cramino takes 12 min with a peak memory usage of 147 Mb without optional output or 21 min with a peak memory usage of 690 Mb for total output including histograms, karyotype, phasing metrics, MD5 checksum, and generation of the Arrow file. Kyber is a tool for creating standardized plots of log-transformed read length versus (phred-scaled) gap-compressed reference identity, for a single input file or a comparison of up to three CRAM or BAM files (Supplementary Methods and Supplementary Fig. S2).

4Phasius is developed to visualize the results of read phasing, which shows in a dynamic genome browser style the length and interruptions between contiguously phased blocks from a large number of individuals together with genome annotation, for example segmental duplications (Supplementary Methods and Fig. 1). Phasius takes 26 s to generate the example figure for 92 individuals in a 10 megabase interval, with eight parallel threads and a peak memory usage of 4.3Gbyte. For the example figure, reads were aligned with minimap2, and alignments phased with longshot.

5NanoPack now offers tools for evaluating large populations with implementations in a more performant programming language, focusing on features relevant to long-read sequencing. The software suite remains easy to install on all major operating systems and offers interactive visualization in HTML format.