Rare variant calls in ultra-deep sequencing

Abstract

Accurate variant calling methods for low frequency variants are provided. Sequence reads of targeted ultra-deep sequencing are received and aligned to a reference sequence. Read depths and variant counts for variants of the same class at each location where the reference allele exists on the reference sequence are determined for each sample-amplicon. Based on the read depths and variant counts, a probability value indicating the confidence level that a specific variant at a specific location is a true positive is calculated using methods such as a statistical model based method and a localized method using a reference sample. The probability value is then compared with a threshold level to determine whether the detected variants are true positives.