281 – Statistical Issues in Evaluation of Therapeutic Devices and Diagnostics
Development of Fast, Slim, and Accurate Amplicon Variant Detection Algorithm for Next-Generation Sequencing
Yan Li
Roche Molecular Systems, Inc.
Wei-Min Liu
Roche Molecular Systems, Inc.
Amplicon variant detection (AVD) is one of the important applications for next generation sequencing (NGS). The enormous amount of sequencing data generated by the NGS technology necessitates the development of fast, slim and accurate AVD algorithm. Here, we present our approaches to develop and implement an AVD algorithm with fast processing speed and efficient use of memory space. The hash function and Burrows-Wheeler transformation are customized to speed up AVD so that processing the large amount of sequencing reads can be completed in a relative short time. We also propose a special way to compress the alignments of reads with reference sequences. Our format uses significantly smaller memory space in comparison with some widely used formats such as SAM/BAM. Moreover, it is useful to automatically identify whether the newly found variants exist in a public database such as COSMIC. We also discuss accuracy and other issues in evaluation of AVD algorithms for NGS.