de novo variant detection

The detection of de novo small variants using the DRAGEN algorithm is performed directly from gVCF files rather than BAM files (as in some other algorithms such as Platypus). In generating gVCF files, homozygous reference variants with similar quality scores from consecutive genomic positions are collapsed into a group and represented by a single entry in the resulting file. Consequently, metrics (e.g., quality scores, depth of coverage, allelic fractions etc) relating to specific sites with homozygous reference genotypes may not be available. Thus, care should be taken when interpreting the apparent allelic depth in parental samples at sites corresponding to de novo variants in their offspring (i.e., homozygous reference positions in the parents) as the metrics presented in VCF files may not correspond to the anticipated position.

This behavior applies to all chromosomes, including sex chromosomes and the mitochondrial genome. Allele counts for homozygous reference positions can be obtained directly from the BAM file, for example by viewing the BAM file in IGV or generating a pile-up using bcftools for a specified genomic position.

The DRAGEN de novo small variant detection algorithm determines all positions for which the genotypes in a trio are not consistent with a Mendelian inheritance pattern. Detection of de novo variants is not restricted to variant positions with homozygous reference genotypes for the parents and a heterozygous genotype for the offspring.