Increased Sensitivity of Diagnostic Mutation Detection by Re-analysis Incorporating Local Reassembly of Sequence...

Diagnostic genetic testing programmes based on next-generation DNA sequencing have resulted in the accrual of large datasets of targeted raw sequence data. Most diagnostic laboratories process these data through an automated variant-calling pipeline. Validation of the chosen analytical methods typically depends on confirming the detection of known sequence variants. Despite improvements in short-read alignment methods, current pipelines are known to be comparatively poor at detecting large insertion/deletion mutations.

We performed clinical validation of a local reassembly tool, ABRA (assembly-based realigner), through retrospective reanalysis of a cohort of more than 2000 hereditary cancer cases.

ABRA enabled detection of a 96-bp deletion, 4-bp insertion mutation in PMS2 that had been initially identified using a comparative read-depth approach. We applied an updated pipeline incorporating ABRA to the entire cohort of 2000 cases and identified one previously undetected pathogenic variant, a 23-bp duplication in PTEN. We demonstrate the effect of read length on the ability to detect insertion/deletion variants by comparing HiSeq2500 (2 × 101-bp) and NextSeq500 (2 × 151-bp) sequence data for a range of variants and thereby show that the limitations of shorter read lengths can be mitigated using appropriate informatics tools.

This work highlights the need for ongoing development of diagnostic pipelines to maximize test sensitivity. We also draw attention to the large differences in computational infrastructure required to perform day-to-day versus large-scale reprocessing tasks.