Comparative evaluation of Oxford Nanopore Technologies' adaptive sampling and the Twist long-read PGx panel for pharmacogenomic profiling

Clinical pharmacogenomics (PGx) testing strategies are mainly based on targeted PCR, microarrays, or short-read sequencing. These methods perform well for detecting known single-nucleotide variants (SNVs), small insertions/deletions (indels), and certain copy number variants (CNVs), but they fall short in resolving complex structural variants (SVs), particularly in complex pharmacogenes such as <i>CYP2D6</i>. Therefore, we previously developed a targeted PGx test based on long-read Oxford Nanopore Technologies (ONT) sequencing. Harnessing adaptive sampling (AS) for <i>in silico</i> enrichment of a panel of PGx genes, we illustrated superior performance in star-allele calling compared to the Genetic Testing Reference Materials Program (GeT-RM) truth set. However, accurate diplotyping of <i>CYP2D6</i> remained challenging. In this work, we adopted the latest basecalling, variant calling, phasing, and star-allele calling tools on our pre-existing data from the HG001, HG01190, NA19785, HG002, and HG005 reference samples. Additionally, we benchmarked the results to public data obtained using the long-read compatible Twist Alliance PGx panel. The re-analyzed ONT-AS data demonstrated correct <i>CYP2D6</i> star-alleles compared to the GeT-RM truth set. Upon benchmarking to the Twist Alliance PGx panel, perfect star-allele matching was obtained between our panel and the Twist PGx panel for all included Clinical Pharmacogenomics Implementation Consortium (CPIC) Level A genes. However, our ONT-AS panel demonstrated superior variant phasing, resulting in three times more variants per phasing block. These findings confirm the robustness of ONT-AS for targeted long-read PGx applications and highlight its potential to support more accurate pharmacogenomic testing, particularly for structurally complex genes like <i>CYP2D6</i>.