pHapCompass: Probabilistic Assembly and Uncertainty Quantification of Polyploid Haplotype Phase

TL;DR

HapCompass, probabilistic haplotype assembly algorithms for diploid and polyploid genomes that explicitly model and propagate read assignment ambiguity to compute a distribution over polyploid haplotype phasings are presented and graph theoretic algorithms to enable statistical inference and uncertainty quantification despite an exponential space of possible phasings are developed.

Abstract

Computing haplotypes from sequencing data, i.e. haplotype assembly, is an important component of molecular and population genetics problems, including interpreting the effects of genetic variation on complex traits and reconstructing genealogical relationships. Assembling the haplotypes of polyploid genomes remains a significant challenge due to the exponential search space of haplotype phasings and read assignment ambiguity; the latter challenge is particularly difficult for haplotype assemblers since the information contained within the observed sequence reads is often insufficient for unambiguous haplotype assignment in polyploid genomes. We present pHapCompass, probabilistic haplotype assembly algorithms for diploid and polyploid genomes that explicitly model and propagate read assignment ambiguity to compute a distribution over polyploid haplotype phasings. We develop graph theoretic algorithms to enable statistical inference and uncertainty quantification despite an exponential space of possible phasings. Since prior work evaluates polyploid haplotype assembly on synthetic genomes that do not reflect the realistic genomic complexity of polyploidy organisms, we develop a computational workflow for simulating genomes and DNA-seq for auto- and allopolyploids. Additionally, we generalize the vector error rate and minimum error correction evaluation criteria for partially phased haplotypes. Benchmarking of pHapCompass and several existing polyploid haplotype assemblers shows that pHapCompass yields competitive performance across varying genomic complexities and polyploid structures while retaining an accurate quantification of phase uncertainty. The source code for pHapCompass, simulation scripts, and datasets are freely available at https://github.com/bayesomicslab/pHapCompass.

Figures (26)

• Figure 1: Haplotype assembly input for a triploid genome. Long (green) or paired-end short (red) reads that cover two or more heterozygous SNPs (encoded as 0 or 1 for major or minor alleles) are informative of haplotype phase. A collection of overlapping reads (e.g., positions 1-4) can be assembled into a single phasing (haplotype block) but their haplotype phase relative to other blocks is undetermined (e.g., positions $L-1$ and $L$).
• Figure 2: Graph constructions for pHapCompass-short inference.Left: The SNP graph $G = (V, E)$ where vertices $v_0, \ldots, v_L$ represent heterozygous SNP positions and edges connect positions covered by at least one sequencing read. Middle: The SNP line graph $Q = (U, E_Q)$, where each node $v_{i,j}$ corresponds to an edge $(i, j)$ in $G$, and two nodes are adjacent if their corresponding edges share exactly one SNP position. Right: The pCompass graph, a factor graph over $Q$ where node potentials $\phi_{v_{i,j}}$ encode the likelihood of phasings for SNP pairs given read evidence, and clique potentials (e.g., $\phi_{v_{0,1}, v_{0,2}, v_{1,2}}$) capture phasing evidence from reads spanning three or more positions.
• Figure 3: pHapCompass-long graphical model. The dashed undirected edges from each $w_\ell^k$ to $W_\ell$ and around each $W_\ell$ represent a deterministic dependence between $w_\ell^k$ and $W_\ell$ (rather than statistical).
• Figure 4: Haplotype assembly performance on simulations. (a-d) VER (top) and MEC (bottom) for pHapCompass (blue), WhatsHap (orange), H-PoPG (green), and HapTree-X (red) on: (a) autopolyploid short reads, (b) autopolyploid long reads, (c) allopolyploid short reads, (d) allopolyploid long reads.
• Figure 5: Uncertainty quantification via FFBS sampling. Mean Hamming distance versus SNP distance for (a) pHapCompass-short and (b) pHapCompass-long across ploidies.