Implementation of polygenic risk scores from sequencing data towards practice by utilizing large publicly available datasets

Abstract

Methods: We developed a protocol for long-read targeted sequencing using capture probes from Twist Bioscience and applied this workflow to sequence 21 pharmacogenes from 41 samples with PacBio HiFi technology. Results: In total, 41 samples had an average on target phasing of 62% (47%-73%) and the average haploblock size was 7,509bp demonstrating the large number of nucleotides in the target region that were phased. In the CYP3A locus, 1,088 unique variants were detected, of which 570 variants were located in the core regions of CYP3A4, CYP3A5 and CYP3A7. Only 27 of these variants (2%) are included in the clinically used *-allele nomenclature. Notably, 1 frameshift-, 5 missense-and 8 splice site variants which are not included in clinical nomenclature were detected. Per individual, an average of 155 unique variants were detected and 34% (5%-86%) of nucleotides were phased in the CYP3A locus. Conclusions: Our results indicate that a panel-based long-read sequencing approach can phase the majority of variants in complex genomic regions, revealing a high abundance of unknown but potentially impactful variants in the CYP3A locus.