A graph-based approach to diploid genome assembly

Abstract

Motivation: Constructing high-quality haplotype-resolved de novo assemblies of diploid genomes is important for revealing the full extent of structural variation and its role in health and disease. Current assembly approaches often collapse the two sequences into one haploid consensus sequence and, therefore, fail to capture the diploid nature of the organism under study. Thus, building an assembler capable of producing accurate and complete diploid assemblies, while being resource-efficient with respect to sequencing costs, is a key challenge to be addressed by the bioinformatics community.

Results: We present a novel graph-based approach to diploid assembly, which combines accurate Illumina data and long-read Pacific Biosciences (PacBio) data. We demonstrate the effectiveness of our method on a pseudo-diploid yeast genome and show that we require as little as 50× coverage Illumina data and 10× PacBio data to generate accurate and complete assemblies. Additionally, we show that our approach has the ability to detect and phase structural variants.

Availability and implementation: https://github.com/whatshap/whatshap.

Supplementary information: Supplementary data are available at Bioinformatics online.

Fig. 1.

Based on reads (middle) from the two sequences (top), the bubbles in the graph (bottom) show three different heterozygous variants; the first one is an SNV, the second one is an SV, and the third one is an indel

Fig. 2.

Input: an assembly graph (top) (consisting of four SNVs and two SVs) and the PacBio reads $r_1,r_2,r_3,r_4,r_5,r_6$ (gray). Output: the phased reads (colored in blue and red) and haplotigs (bottom) using Falcon Unzip and our approach. Our graph-based approach also phases the central region. Contrarily, Falcon Unzip does not phase it, and so the region does not contribute to the total haplotig size

Fig. 3.

Overview of the diploid assembly pipeline

Fig. 4.

For a subgraph of G_s, the example shows two bubbles l₁ and l₂, and their corresponding alleles. Reads $r_1,r_2,r_3,r_4$ traverse these bubbles

Fig. 5.

For a subgraph of G_s, this example shows the true (top) and predicted (bottom) versions of two haplotype alignments (red and blue) through a series of bubbles. When comparing the correspondingly-colored lines between the two versions, we see one switch between SV1 and SV2: the prediction contains one switch error. Six bubbles have been phased, for a total of five phase connections between consecutive bubbles. Therefore, the phasing error rate is 1/5

Fig. 6.

Structural variation analysis of phased bubbles from our graph-based approach. (a) Joint distribution of allele length and Hamming distance, for pure substitutions. (b) Distribution of size difference between the two alleles, for mixed bubbles and indels. Pure substitutions always have a size difference of 0, and are not included in the figure. (c) Joint distribution of the length of the longer allele and the substitution rate, for mixed bubbles. With a higher substitution rate, the bubble has more substitutions, and with a lower rate more indels