Back

Targeted adaptive sampling enables clinical pharmacogenomics testing and genome-wide genotyping

Gan Hui Peng, P.; Han Lin, Y.; Irfan Bin Hajis, M.; Maulana, Y.; Ng Qi Hui, A.; Nathanael Ramanto, K.; Nisita Dewanggana, M.; Irwanto, A.; Sani, L.; Goh Liuh, L.; Gonzalez-Porta, M.

2025-05-07 pharmacology and therapeutics
10.1101/2025.05.05.25326970 medRxiv
Show abstract

Pharmacogenomics (PGx) testing improves medication safety and efficacy by identifying genetic variants that affect drug response. However, current technologies often fail to resolve complex loci, detect structural variants, or phase alleles accurately. Here, we present an end-to-end PGx workflow based on Targeted Adaptive Sampling-Long Read Sequencing (TAS-LRS), integrating a streamlined laboratory protocol with a bioinformatics pipeline that includes a novel CYP2D6 caller. Using 1,000 ng of DNA and three-sample multiplexing on a single PromethION flow cell, the assay achieves consistent on-target (25x) and off-target (3x) coverage, enabling accurate, haplotype-resolved testing of 35 pharmacogenes alongside genome-wide genotyping from off-target reads. We further developed the workflow into a clinically ready service and validated its performance across 17 reference and clinical samples. The assay demonstrated high concordance for small variants (99.9%) and structural variants (>95%), with phased diplotypes and metabolizer phenotypes reaching 97.7% and 98.0% concordance, respectively. Improved calls were observed in 12 genes due to enhanced genotyping, phasing, or novel allele detection. In addition, off-target reads supported accurate genome-wide imputation, comparable to short-read sequencing and superior to microarrays. These results establish the feasibility of long-read sequencing for clinical PGx testing and position TAS-LRS as a scalable solution combining both targeted and genome-wide utility.

Matching journals

The top 5 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.