An Algorithm for Estimating Contig Copy Number for Polyploid Genomes

There are currently many available methodologies for de novo assembly of entire genomes from short reads obtained by whole genome shotgun (WGS) sequencing, and many diploid organisms have been successfully sequenced to date. A few autopolyploid organisms have also been sequenced, although through more time and resource consuming strategies such as BAC sequencing or doubled monoploids, which may not be applicable to all species, particularly one with high ploidy. When applied to WGS data from polyploids, diploid oriented assemblers collapse multiple chromosome copies into the same contig, such that the final assembly is a mixture of completely haplotyped and coalesced segments, with varying degrees of collapse. We will describe an algorithm for ploidy estimation of contigs and scaffolds from de novo assemblies of polyploid species, based on high throughput next generation sequencing data. This method leverages the relative proportions of both alleles for heterozygous positions, while controlling for sequencing errors via a learned error model that takes into account systematic errors produced by sequencing technologies. Scaffolds with ploidy ranging from 1 to 16 were simulated and results showed that 15X coverage per haploid segment was sufficient for accurate ploidy estimation. Real diploid human data was used for further validation and correct results were obtained. This algorithm also provides a natural framework for SNP calling.  Using the approach, we have obtained lower false positive and negative rates than the most widely used variant caller.