Haploid to diploid alignment for variation calling assessment

Abstract

MotivationVariation calling is the process of detecting differences between donor and consensus DNA via high-throughput sequencing read mapping. When evaluating the performance of different variation calling methods, a typical scenario is to simulate artificial (diploid) genomes and sample reads from those. After variation calling, one can then compute precision and recall statistics. This works reliably on SNPs but on larger indels there is the problem of invariance: a predicted deletion/insertion can differ slightly from the true one, yet both make the same change to the genome. Also exactly correct predictions are rare, especially on larger insertions, so one should consider some notion of approximate predictions for fair comparison.ResultsWe propose a full genome alignment-based strategy that allows for fair comparison of variation calling predictions: First, we apply the predicted variations to the consensus genome to create as many haploid genomes as are necessary to explain the variations. Second, we align the haploid genomes to the (aligned) artificial diploid genomes allowing arbitrary recombinations. The resulting haploid to diploid alignments tells how much the predictions differ from the true ones, solving the invariance issues in direct variation comparison. In an effort to make the approach scalable to real genomes, we develop a simple variant of the classical edit distance dynamic programming algorithm and apply the diagonal doubling technique to optimise the computation. We experiment with the approach on simulated predictions and also on real prediction data from a variation calling challenge.

Repeat- and error-aware comparison of deletions

“…The alignment score gives a ranking for deletion prediction in the case that one of the sets corresponds to the ground truth. This approach was recently extended to the diploid ground-truth setting (Mäkinen and Rahkola, 2013) with an O(dn) algorithm, where d is the unit cost edit distance between predicted sequence and ground-truth, and n is the maximum sequence length. In practice, this approach is not scalable to whole-genome comparisons (Mäkinen and Rahkola, 2013); computation took 6 to 11.5 hours, depending on prediction accuracy, only for the 63 Mbp long human chromosome 20.…”

Recombination-aware alignment of diploid individuals

“…Another application is possible in variant calling evaluation [7], as covering alignment takes heterozygous variations properly into account: High-throughput sequencing allows a cost-effective way to discover how an individual genome differs from the consensus reference genome of the species. The result of such variant calling process is a list of homozygous and heterozygous variant predictions.…”

“…Because of that, in [7] all predicted variants are applied to the reference in order to create a predicted haploid genome . Edit distance between the predicted haploid and the ground-truth diploid was then computed, allowing arbitrary recombinations for the haploid to distribute along the diploid, giving a distance measure.…”

“…Edit distance between the predicted haploid and the ground-truth diploid was then computed, allowing arbitrary recombinations for the haploid to distribute along the diploid, giving a distance measure. It was shown that this haploid to diploid edit distance can be computed on realistic size variant calling scenarios [7]. This approach is actually highly similar to the path-alignments of [5,6].…”

“…There remained one shortcoming in [7] due to the asymmetric measure; if there are overlapping predicted heterozygous variants, one needs to decide which ones to take to the predicted haploid. It was proposed to create another haploid with the remaining predictions that could not be applied in the first round, and do another haploid to diploid distance computation.…”