611 – Analysis of Diagnostic Sequencing and Related Data
Statistical Scores for Rare Variant Calls in Ultra-Deep Sequencing
Wei-Min Liu
Roche Molecular Systems
NGS technology is changing bio-medical sciences from academic research to clinical diagnostics. For non-invasive blood-based tests, it is crucial to distinguish the rare variants from sequencing noise in ultra-deep sequencing. Several approaches have been developed to make the variant calls reliable. They include base call quality scores, unique molecular identifiers, etc. Most software packages only call variants of 1% or higher by their default setting to avoid false positives. I describe the new variant quality scores based on the distribution of false positives in sequencing, as well as the fact that the false positive rates are dependent on the sequence contexts and locations. With statistical tests based on these considerations, we can detect variants with percentages significantly below 1% (depending on variants, sample types and sequencing protocols) when sufficient number of DNA molecules is present in the sample. I also describe a non-standard usage of MiSeq Reporter (MSR) to verify the low-frequency variants we found using this approach.