Validation of a comprehensive long-read sequencing platform for broad clinical genetic diagnosis

Siddhartha Sen^1*Hillary P Handler²Alec Victorsen¹Zach Flaten¹Aidan Ellison¹Todd Knutson³Sarah Munro³Ryan J Martinez¹Charles John Billington¹Jennifer J Laffin¹Sarah Bray²Pawel Mroz¹Sophia Yohe¹Andrew C Nelson¹Matthew Bower²Bharat Thyagarajan^1*

• ¹University of Minnesota Health Sciences, University of Minnesota Medical Center, Minneapolis, United States
• ²Fairview Health Services, Minneapolis, Minnesota, United States
• ³University of Minnesota Twin Cities, St. Paul, Minnesota, United States

Though short read high-throughput sequencing, commonly known as Next Generation Sequencing (NGS), has revolutionized genomics and genetic testing, there is no single genetic test that can accurately detect single nucleotide variants (SNVs), small insertions/deletions (indels), complex structural variants (SVs), repetitive genomic alterations, and variants in genes with highly homologous pseudogenes. The implementation of a unified comprehensive technique that can simultaneously detect a broad spectrum of genetic variation would substantially increase efficiency of the diagnostic process. The current study evaluated the clinical utility of long-read sequencing as a comprehensive genetic test for diagnosis of inherited conditions. Using Oxford Nanopore Technology (ONT) long read nanopore sequencing, we successfully developed and validated a clinically deployable integrated bioinformatics pipeline that utilizes a combination of eight publicly available variant callers. A concordance assessment comparing the known variant calls from a well-characterized, benchmarked sample called NA12878 from the National Institute of Standards and Technology (NIST) with the variants detected by our pipeline for this sample, determined that the analytical sensitivity of our pipeline was 98.87% and the analytical specificity exceeded 99.99%. We then evaluated our pipeline's ability to detect 167 clinically relevant variants from 72 clinical samples. This set of variants consisted of 80 SNVs, 26 indels, 32 SVs, and 29 repeat expansions, including 14 variants in genes with highly homologous pseudogenes. The overall detection concordance for these clinically relevant variants was 99.4% (95% CI: 99.7%-99.9%). Importantly, in addition to detecting known clinically relevant variants, in four cases, our pipeline yielded valuable additional information in support of clinical diagnoses that could not have been established using short-read NGS alone. Our findings suggest that long-read sequencing is successful in identifying diverse genomic alterations and that our pipeline functions well as the basis for a single diagnostic test for patients with suspected genetic disease.