HiCanu: accurate assembly of segmental duplications, satellites, and allelic variants from high-fidelity long reads

Abstract

Complete and accurate genome assemblies form the basis of most downstream genomic analyses and are of critical importance. Recent genome assembly projects have relied on a combination of noisy long-read sequencing and accurate short-read sequencing, with the former offering greater assembly continuity and the latter providing higher consensus accuracy. The recently introduced Pacific Biosciences (PacBio) HiFi sequencing technology bridges this divide by delivering long reads (>10 kbp) with high per-base accuracy (>99.9%). Here we present HiCanu, a modification of the Canu assembler designed to leverage the full potential of HiFi reads via homopolymer compression, overlap-based error correction, and aggressive false overlap filtering. We benchmark HiCanu with a focus on the recovery of haplotype diversity, major histocompatibility complex (MHC) variants, satellite DNAs, and segmental duplications. For diploid human genomes sequenced to 30× HiFi coverage, HiCanu achieved superior accuracy and allele recovery compared to the current state of the art. On the effectively haploid CHM13 human cell line, HiCanu achieved an NG50 contig size of 77 Mbp with a per-base consensus accuracy of 99.999% (QV50), surpassing recent assemblies of high-coverage, ultralong Oxford Nanopore Technologies (ONT) reads in terms of both accuracy and continuity. This HiCanu assembly correctly resolves 337 out of 341 validation BACs sampled from known segmental duplications and provides the first preliminary assemblies of nine complete human centromeric regions. Although gaps and errors still remain within the most challenging regions of the genome, these results represent a significant advance toward the complete assembly of human genomes.

Figure 1.

Impact of HiCanu processing on observed read quality. (A) Two hypothetical reads are shown with sequencing errors highlighted in red. (B) The first step of HiCanu is to compress homopolymers, which obscures homopolymer length errors but retains enough information to accurately distinguish reads from different genomic loci. (C) Overlaps are then computed for the compressed reads, and remaining errors are identified by examining the alignment pileups (gray rectangle). (D) Finally, after correcting the identified errors (blue) and ignoring indels in regions of known systematic error (gray), the resulting overlap is 100% identical. (Right) Sequence identity of reads from a 20-kbp HiFi library measured against the CHM13 Chromosome X reference sequence v0.7 (Miga et al. 2020) after each step of HiCanu processing (Supplemental Note 1). Separate boxplots are shown for initial raw HiFi reads (init), homopolymer-compressed reads (compressed), OEA-corrected reads (corrected), and corrected reads after ignoring differences in microsatellite repeats (masked). The median read identity, indicated by solid segments, increases from <99.9% to 100% (note the plot shows y-range of 99.65%–100%). Supplemental Table S1 also shows how HiCanu processing increases the percentage of perfectly aligned (100% identity) HiFi reads from <1% to >97%.

Figure 2.

Visual representation of the most continuous HiFi-based and Nanopore-based assemblies of the CHM13 genome. HiCanu assembly of the 20-kbp HiFi data set (left) and Canu assembly of an ultralong Nanopore data set (right). White regions indicate gaps in the current reference genome, and each gray and black block indicates a continuous contig alignment. Color switches from gray to black represent either the end of a contig or an alignment break. Assemblies were aligned to GRCh38 using MashMap (Jain et al. 2018a), and plots were generated using coloredChromosomes (Böhringer et al. 2002) as previously described (Berlin et al. 2015; Jain et al. 2018b). Note that some chromosomes (e.g., Chr X) are better resolved by the Nanopore assembly owing to the presence of near-perfect repeats. At the same time, chromosomes containing more diverged repeats (e.g., Chr 7 and Chr 16) are better resolved by the HiFi assembly. We note that some gaps in the HiFi assembly are caused by sequence-specific biases of current HiFi sequencing protocols (Supplemental Note 4). The red box highlights the defensin beta gene family on Chromosome 8p23.1 which is split in both assemblies and detailed in Figure 4.

Figure 3.

HiCanu assembly of the CHM13 Chromosome 19 centromere. RepeatMasker (Smit et al. 2013) of tig00006497 reveals three α-satellite HOR arrays that reside within the Chromosome 19 centromere (D19Z1, D19Z2?, and D19Z3; marked with black bars). These HOR arrays are 606 kbp, 289 kbp, and 3.96 Mbp in length, respectively, and are composed of a 13-mer, a complex higher-order HOR, and a dimeric HOR unit, respectively. The HOR repeat underlying D19Z2 shares limited sequence identity with the pG-A16 repeat previously described (Hulsebos et al. 1988; Choo et al. 1991; Finelli et al. 1996) and, therefore, is designated with a question mark. The α-satellite HOR arrays have relatively uniform coverage of HiFi and ultralong Oxford Nanopore data, except for a drop in Oxford Nanopore sequencing coverage over the D19Z1 array, which may be owing to a misassembly, read mismapping, or biases in sequencing. The HiFi coverage plot shows fold coverage of the most common base (black) and the second most common base (red).

Figure 4.

Chr 8 defensin beta cluster repeat structure and assembly comparison. (Top) NUCmer self-alignment dot plots (Kurtz et al. 2004) of the CHM13 reference defensin beta cluster at different alignment stringencies (Methods): (A) >7 kbp repeats at 98% identity. (B) >7 kbp repeats at 99.9% identity. Purple/blue indicates same/reverse strand matches. (C) Icarus (Mikheenko et al. 2016) visualization of contig alignments from both HiFi-based (Canu, HiCanu, Peregrine) and ultralong Nanopore-based assemblies (Canu ONT and Flye ONT) (Kolmogorov et al. 2019) produced by QUAST (Gurevich et al. 2013). White space in the alignment figure indicates the assembly was fragmented into short contigs (<50 kbp). Red color indicates misassembled contigs. The HiCanu assembly breaks at two of three SD instances that share high sequence similarity (black arrows) and at a region of systematic HiFi coverage depletion (red arrow).