Software-assisted manual review of clinical NGS data:an alternative to routine Sanger sequencing confirmationwith equivalent results in >15,000 hereditary cancer screens
Clinical genomic tests increasingly utilize a next generation sequencing (NGS) platform due in part to the high fidelity of variant calls, yet rare errors are still possible. In hereditary cancer screening, failure to correct such errors could have serious consequences for patients, who may follow an unwarranted screening or surgical-management path. It has been suggested that routine orthogonal confirmation via Sanger sequencing is required to verify NGS results, especially low-confidence positives with depressed allele fraction (<30% of alternate allele). We evaluated whether an alternative method of confirmation--software-assisted manual call review--performed comparably to Sanger confirmation in >15,000 samples. Licensed reviewers manually inspected both raw and processed data at the batch-, sample-, and variant-level, including raw NGS read pileups. Of ambiguous variant calls with <30% allele fraction (1,707 total calls at 38 unique sites), manual call review classified >99% (1,701) as true positives (enriched for long insertions or deletions (\"indels\") and homopolymers) or true negatives (often conspicuous NGS artifacts), with the remaining <1% (6) being mosaic. Critically, results from software-assisted manual review and retrospective Sanger sequencing were concordant for samples selected from all ambiguous sites. We conclude that the confirmation required for high confidence in NGS-based germline testing can manifest in different ways: a trained NGS expert operating platform-tailored review software achieves quality comparable to routine Sanger confirmation.