Allele Identification in Assembled Genomic Sequence Datasets

Abstract

Allelic variation within species provides fundamental insights into the evolution and ecology of organisms, and information about this variation is becoming increasingly available in sequence datasets of multiple and/or outbred individuals. Unfortunately, identifying true allelic variants poses a number of challenges, given the presence of both sequencing errors and alleles from other closely related loci. We outline the key considerations involved in this process, including assessing the accuracy of allele resolution in sequence assembly, clustering of alleles within and among individuals, and identifying clusters that are most likely to correspond to true allelic variants of a single locus. Our focus is particularly on the case where alleles must be identified without a fully resolved reference genome, and where sequence depth information cannot be used to infer the putative number of loci sharing a sequence, such as in transcriptome or post-assembly datasets. Throughout, we provide information about publicly available tools to aid allele identification in such cases.

Allele Identification for Transcriptome-Based Population Genomics in the Invasive PlantCentaurea solstitialis

“…This low SNP density underscores the need for long-read approaches to accurately recover haplotypes within and among conspecific individuals (Dlugosch and Bonin 2012; Lai et al 2012). The proportion of these sites sequenced in each individual increased sharply with sequencing effort (Figure 3A), with no indication of saturation; the largest libraries covered less than 60% of SNP positions.…”

“…AllelePipe was used to verify that similar sequences aligned throughout their region of overlap (expected for true alleles), and to cluster groups of similar sequences via single-linkage clustering. Single-linkage clustering generates maximal aggregation of sequences, and will bring together both closely related paralogs and their alleles (Dlugosch and Bonin 2012). Multiple alignments were created for sequences within each cluster and their consensus sequence generated using CAP3.…”