HMMploidy: inference of ploidy levels from short-read sequencing data

Abstract

The inference of ploidy levels from genomic data is important to understand molecular mechanisms underpinning genome evolution. However, current methods based on allele frequency and sequencing depth variation do not have power to infer ploidy levels at low-and mid-depth sequencing data, as they do not account for data uncertainty. Here we introduce HMMploidy, a novel tool that leverages the information from multiple samples and combines the information from sequencing depth and genotype likelihoods. We demonstrate that HMMploidy outperforms existing methods in most tested scenarios, especially at low-depth with large sample size. We apply HMMploidy to sequencing data from the pathogenic fungus Cryptococcus neoformans and retrieve pervasive patterns of aneuploidy, even when artificially downsampling the sequencing data. We envisage that HMMploidy will have wide applicability to low-depth sequencing data from polyploid and aneuploid species.

Hybridisation and chloroplast capture between distinct Themeda triandra lineages in Australia

“…The ploidy of each sample was estimated using HMMploidy (Soraggi et al, 2021 ), a method which has been developed to infer ploidy from low‐depth sequencing data. HMMploidy uses both sequencing depth and genotype likelihoods to infer ploidy, leveraging population frequencies to account for genotype uncertainty in low‐coverage data (Soraggi et al, 2021 ).…”

ngsJulia: population genetic analysis of next-generation DNA sequencing data with Julia language

“…Nucleotide, genotype and allele frequency likelihoods ngsJulia provides utilities to calculate nucleotide and genotype likelihoods, i.e., the probability of observed sequencing data given a specific nucleotide or genotype, 21 for an arbitrary ploidy level, as in Soraggi et al 22 We now describe how such quantities are calculated in ngsJulia.…”

“…Following the notation in Soraggi et al, 22 for one sample and one site, we let O be the observed NGS data, Y the ploidy, and G the genotype. Therefore, G has values in 0, 1,…, Y f g , i.e., the number of derived (or alternate) alleles.…”

“…In the simplest form, genotype likelihoods can be calculated by considering individual base qualities as probabilities of observing an incorrect nucleotide. 21 We adopt the calculation of genotype likelihoods for an arbitrary ploidy level P OjG, Y ð Þas proposed in Soraggi et al 22 From the genotype likelihoods with P OjG,Y ¼ 1 ð Þ , the two most likely alleles are identified by sorting P OjG, Y ¼ 1 ð Þvalues after pooling all sequencing reads together across all samples. This operation will restrict the range of possible genotypes to biallelic variation only.…”

“…ngsPloidy: analysis of sequencing data from polyploid genomes We further utilised ngsJulia to implement an additional program, ngsPloidy, for the estimation of ploidy from unknown genotypes. The method implemented is similar to the one proposed by Soraggi et al 22 with some notable differences on the calculation of genotype probabilities (see Methods). Additionally, ngsPloidy includes a novel method to test for multiploidy in the sample.…”