Detection of Rare Mutations Using Beta-Binomial and Empirical Quantile Models in Next-Generation Sequencing Experiments

Abstract

Next-generation sequencing is often used to identify genetic variants. The probability of variant detection also depends on the variant caller. Pooled data could be used to lower the sequencing cost. However identifying variants from pooled data is more challenging and demands more sophisticated mathematical methods. In this article we propose two novel SNP calling approaches: modification of Beta-binomial model as proposed in Flaherty et al. (2011) using posterior Beta distribution and empirical quantile method. Both offered methods and original Beta-binomial model were applied to pooled exome data of patients with neuromuscular diseases. The results showed that Beta-binomial model and modification of it were highly specific, however, with lower sensitivity compared to empirical quantile model. The positions could be identified as mutated using empirical quantile model much faster rather Beta-binomial models.