Approaches to long-read sequencing in a clinical setting to improve diagnostic rate

Abstract

Over the past decade, advances in genetic testing, particularly the advent of next-generation sequencing, have led to a paradigm shift in the diagnosis of molecular diseases and disorders. Despite our present collective ability to interrogate more than 90% of the human genome, portions of the genome have eluded us, resulting in stagnation of diagnostic yield with existing methodologies. Here we show how application of a new technology, long-read sequencing, has the potential to improve molecular diagnostic rates. Whole genome sequencing by long reads was able to cover 98% of next-generation sequencing dead zones, which are areas of the genome that are not interpretable by conventional industry-standard short-read sequencing. Through the ability of long-read sequencing to unambiguously call variants in these regions, we discovered an immunodeficiency due to a variant in IKBKG in a subject who had previously received a negative genome sequencing result. Additionally, we demonstrate the ability of long-read sequencing to detect small variants on par with short-read sequencing, its superior performance in identifying structural variants, and thirdly, its capacity to determine genomic methylation defects in native DNA. Though the latter technical abilities have been demonstrated, we demonstrate the clinical application of this technology to successfully identify multiple types of variants using a single test.

Figure 1

Biallelic hypermethylation on chr15 in Prader–Willi syndrome (PWS). Methylation analysis of HiFi reads shows hypermethylation of both haplotypes at known chr15 imprinted loci in a male patient, C5, with Prader–Willi Syndrome. An unrelated, unaffected male control, F12, shows hypomethylation of one allele. HiFi reads are phased by sequence into haplotype 1s and 2. Values show the percent of reads from each haplotype that are methylated at each genomic CpG site. h1 haplotype 1, h2 haplotype 2.

Figure 2

Illustrative schematic of determining HPO terms best assayable by LRS. (A) SRS genomic coverage (gray bars) averaged across hundreds of genomes is calculated for each gene (blue lines). (B) Disease genes are mapped to HPO terms. (C) Terms are assembled and (D) the number of genes and the SRS-uncoverable size are assembled for each HPO term. These can then be used to prioritize patients for long read sequencing.