Variant Call Differences Between Paired Reads in Illumina Amplicon Sequencing: An Issue for Rare Mutation Detection

• Activity Number: 471 - Statistical Methods and Challenges in Diagnostic Medicine
• Type: Contributed
• Date/Time: Wednesday, August 2, 2017 : 8:30 AM to 10:20 AM
• Sponsor: Section on Medical Devices and Diagnostics
• Author(s): Wei-Min Liu* and Julie Chen
• Companies: Roche Molecular Systems Inc. and UC Berkeley
• Keywords: FFPET ; liquid biopsy ; mutation detection ; deep amplicon sequencing

Abstract

Deep amplicon sequencing is crucial for finding rare somatic mutations from liquid biopsy or formalin-fixed paraffin-embedded tissues. It can be used as a reference method for developing PCR assay or used directly as diagnosis assay for individualized health care based on mutation detection. It is important to extract actual mutation signals from sequencing noise. Using the RMS proprietary software SWISSA, we studied a special type of sequencing noise appearing in one of the paired reads but not both in Illumina amplicon sequencing. We found that the variants occurring in read 2 but not read 1 are often higher than those occurring in read 1 but not read 2. We also calculated the proportion of this type of noise in all observed variants. We compared the results of using paired-end reads and using only single-end reads for mutation frequencies in a dataset of dilution experiments. We found that the two methods generated similar results and using paired-end reads with removal of this type of sequencing noise may be slightly better than using single-end reads. Our proposed metrics R1Not2 and R2Not1 in VPKH could be used to optimize the experimental conditions in assay development.