DeNovoGear: de novo indel and point mutation discovery and phasing

Abstract

We present DeNovoGear software for analyzing de novo mutations from familial and somatic tissue sequencing data. DeNovoGear uses likelihood-based error modeling to reduce the false positive rate of mutation discovery in exome analysis and fragment information to identify the parental origin of germ-line mutations. We used DeNovoGear on human whole-genome sequencing data to produce a set of predicted de novo insertion and/or deletion (indel) mutations with a 95% validation rate.

Figure 1

Using beta-binomial (‘BB’) likelihoods to model exome data. A) We fit three functions to the read count data from the exomes of NA12878, NA12891, and NA12892, considering only high-confidence SNP sites. For each function, we plot the expected variance in the number of alternate alleles sampled as a function of read depth. One the same scale we show the observed variance for all three exome datasets. B) We performed the same analysis using indel sites and observed an even larger difference in fit between beta-binomial and binomial models than with SNPs. When we examine mutant allele frequencies in the CEU WES dataset at the sites called by using binomial (n=153), beta-binomial (n=66), or both models (n=40), we see that the BB model primarily reduces false positives by eliminating undercalling of heterozygotes in the parents. (C) Distribution of mutant allele frequencies for previously validated sites. On the x-axis, sites are positioned by the cumulative mutant read frequency in the parents. On the y-axis sites are positioned by the mutant read frequency in the trio offspring. Points are colored by validation status. (D) Distribution of mutant allele frequencies for sites called in this study and not validated previously. Sites are colored green if called only by binomial model, red if called only by BB, blue if called both models.