---
title: "Pharmacogenomic landscape in Thailand: Array-based PGx profiling and EMR-linked medication exposure"
id: "plos-one-9-pharmacogenomic-landscape-in-thailand-array-based-profiling-and-emr-linked"
canonical_url: "https://medichelpline.com/clinical-feed/plos-one-9-pharmacogenomic-landscape-in-thailand-array-based-profiling-and-emr-linked"
content_type: "clinical_feed_article"
specialty: "Pharmacology"
source_name: "PLOS ONE (Medicine)"
source_url: "https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0355201"
published_at: "2026-08-03T14:00:00.000Z"
evidence_level: "Journal Feed"
license: "CC-BY-NC-4.0 / Informational Use"
---
# Pharmacogenomic landscape in Thailand: Array-based PGx profiling and EMR-linked medication exposure
## Provenance & Clinical Metadata
- **Canonical URL:** https://medichelpline.com/clinical-feed/plos-one-9-pharmacogenomic-landscape-in-thailand-array-based-profiling-and-emr-linked
- **Specialty:** [Pharmacology](https://medichelpline.com/clinical-feed/pharmacology.md)
- **Primary Source:** PLOS ONE (Medicine)
- **Source URL:** [Original Journal Publication](https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0355201)
- **Published At:** 2026-08-03T14:00:00.000Z
- **Evidence Rating:** Journal Feed
## Executive GIST (TL;DR)
- This study profiled **4,662 Thai adults** using an Asian-optimized SNP array (Infinium Asian Screening Array v1.0) and a pre-specified pharmacogenomic panel (11 genes; 26 markers) aligned to CPIC/PharmVar definitions. - Overall genotype-to-phenotype **callability** across the panel was high at **98.62%**, with most genes >99% callable; **CYP2C19** callability was 95.99% and **NUDT15** 90.28%. - Using phenotype-coded results across nine genes, **95.99% of individuals** carried at least one CPIC-actionable result (median 2 actionable phenotypes per person, IQR 2–3). - Gene-level actionable prevalence was highest for **CYP3A5** (58.54%) and **CYP2C19** (56.67%), then **ABCG2** (45.10%) and **UGT1A1** (27.37%). - EMR linkage to hospital prescription/dispensation data identified **1,529 participants (32.58%)** exposed to one or more study medications; the most commonly used were **omeprazole (n = 658)** and statins (**atorvastatin n = 606; simvastatin n = 603**). - Among medication users, **realized actionability** was notable: **CYP2C19–omeprazole** actionable phenotypes were present in **55.02%** of omeprazole users; **SLCO1B1–statin** actionable phenotypes occurred in **21.95–23.05%** of statin users. - The authors conclude that an **Asian-optimized SNP array** can support scalable PGx phenotyping in Thai adults, and that EMR linkage quantifies realized actionability to identify high-yield targets—particularly **CYP2C19** for proton pump inhibitors and **SLCO1B1** for statins—for pre-emptive implementation. - Study ethics approvals, data access restrictions, and that individual-level genotype and EMR data cannot be shared publicly because of privacy; de-identified minimal data may be requested through the Siriraj Health Study under institutional review. - The work used CPIC level A/B gene–drug relationships to quantify clinical relevance and emphasized integration of population allele frequencies with local prescribing patterns to inform implementation priorities in Thailand.
## Clinical Analysis & Structured Key Points
Pharmacogenomic landscape in Thailand: Array-based profiling and EMR-linked medication exposure | PLOS One Browse Subject Areas ? Click through the PLOS taxonomy to find articles in your field. For more information about PLOS Subject Areas, click here . Article Authors Metrics Comments Media Coverage Reader Comments Figures Figures Abstract Pharmacogenomic (PGx) data in Thailand remain limited, and genetics-only surveys rarely quantify “realized actionability”—the overlap between actionable PGx phenotypes and real-world medication exposure. We profiled 4,662 Thai adults using SNP-array data and a pre-specified PGx panel (11 genes; 26 markers) with a hybrid required/optional calling policy for diplotype/phenotype assignment. CPIC level A/B gene–drug relationships were linked to hospital electronic medical record (EMR) prescription/dispensation data to quantify drug-specific realized actionability. Overall callability across gene-results was 98.62%, exceeding 99% for most genes and lower for CYP2C19 (95.99%) and NUDT15 (90.28%). Across nine phenotype-coded genes, 95.99% carried ≥1 CPIC-actionable result (median 2; IQR 2–3). Actionable prevalence among callable individuals was highest for CYP3A5 (58.54%) and CYP2C19 (56.67%), followed by ABCG2 (45.10%) and UGT1A1 (27.37%). EMR linkage identified 1,529 (32.58%) participants exposed to ≥1 study medication; omeprazole (n = 658) and statins were most common (atorvastatin n = 606; simvastatin n = 603). Among users, actionable phenotypes were frequent for CYP2C19 –omeprazole (55.02%) and SLCO1B1 –statins (21.95–23.05%). In conclusion, an Asian-optimized SNP array supports scalable PGx phenotyping in Thai adults. EMR linkage quantifies realized actionability and highlights high-yield targets ( CYP2C19 –proton pump inhibitors; SLCO1B1 –statins) for pre-emptive implementation. Citation: Pasookhush P, Suta S, Pumeiam S, Mongkolsucharitkul P, Pinsawas B, Ophakas S, et al. (2026) Pharmacogenomic landscape in Thailand: Array-based profiling and EMR-linked medication exposure. PLoS One 21(8): e0355201. https://doi.org/10.1371/journal.pone.0355201 Editor: Nancy Monroy-Jaramillo, INNN: Instituto Nacional de Neurologia y Neurocirugia Manuel Velasco Suarez, MEXICO Received: April 21, 2026; Accepted: July 17, 2026; Published: August 3, 2026 Copyright: © 2026 Pasookhush et al. This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability: All summary data supporting the findings of this study are included in the manuscript and its Supporting Information files. Due to institutional ethics and privacy restrictions, the individual-level genotype and electronic medical record data cannot be shared publicly because they contain potentially identifiable human participant information. Requests for access to de-identified minimal data may be submitted to the Siriraj Health Study at sihealthstudy@mahidol.ac.th . Requests will be reviewed in accordance with participant consent, institutional requirements, and approval by the Institutional Review Board of the Faculty of Medicine Siriraj Hospital, Mahidol University. The de-identified study data will be stored and maintained by the Siriraj Health Study to ensure long-term availability to qualified researchers. Funding: Phongthana Pasookhush (PP) was partially supported by the Faculty of Medicine Siriraj Hospital, Mahidol University. The funder had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing interests: The authors have declared that no competing interests exist. Introduction Inter-individual variability in drug response is a common challenge in clinical practice. Patients treated with the same medication at standard doses may experience different levels of efficacy, toxicity, or both [ 1 , 2 ]. This variability is influenced by many factors, including age, sex, comorbidities, concomitant medications, environmental exposures, and biological determinants such as genetic and epigenetic variation [ 3 ]. Pharmacogenomics (PGx) offers a practical approach to precision therapeutics by linking inherited variation in pharmacogenes to predictable differences in drug metabolism, transport, and exposure [ 4 ]. In this way, PGx can help guide more individualized drug selection and dosing, with the aim of reducing preventable adverse drug reactions and improving treatment benefit [ 5 ]. However, the clinical value of PGx depends not only on genotyping but also on standardized interpretation and prescribing guidance. The Clinical Pharmacogenetics Implementation Consortium (CPIC) publishes peer-reviewed guidelines that translate genotype results into prescribing recommendations (e.g., dose adjustment or alternative therapy) [ 6 , 7 ]. The Pharmacogene Variation (PharmVar) consortium complements this effort by standardizing star-allele nomenclature, which supports consistent genotype-to-phenotype translation and harmonized reporting across laboratories and studies [ 8 , 9 ]. Large-scale biobank studies have demonstrated that clinically actionable PGx variation is common and that many individuals are exposed to medications with established guideline-based recommendations, supporting the potential value of pre-emptive PGx testing at the population level [ 10 ]. However, implementation priorities cannot be assumed to be the same across populations. Allele and phenotype distributions vary across ancestries, and extrapolating frequency estimates from predominantly European datasets may be misleading [ 11 ]. In addition, differences in disease burden and prescribing patterns influence which gene–drug pairs are most commonly encountered in routine care. As a result, implementation priorities depend on local medication utilization patterns as well as genetics [ 11 ]. In Thailand, a whole-genome sequencing (WGS) study has provided an important baseline overview of pharmacogenomic variation [ 12 ]. Nonetheless, genetics-only surveys do not quantify “realized actionability”, defined here as the intersection between actionable PGx phenotypes and observed medication exposure within a healthcare system. They also do not directly address how well scalable genotyping approaches can support implementation-facing PGx reporting. To address these gaps, we performed an array-based PGx analysis in two Thai adult cohorts linked to hospital care and curated a pre-specified panel of clinically relevant pharmacogenes captured on the Infinium Asian Screening Array (ASA), using CPIC/PharmVar-aligned definitions. Our primary objective was to characterize gene-level callability and the distribution of PGx diplotypes and phenotypes in this cohort. Our secondary objective was to link CPIC level A/B gene–drug relationships to routinely collected electronic medical record (EMR) medication data to quantify realized actionability in routine care. By integrating population PGx frequencies with medication exposure in a real-world hospital setting, this study aims to inform implementation priorities by identifying high-frequency gene–drug pairs with high expected clinical reach for pre-emptive screening and guideline-driven prescribing workflows in Thailand. Materials and methods Study cohort and genotyping We conducted a secondary analysis of two prospective studies: the Siriraj Health (SIH) [ 13 ] study and the Siriraj OneHealth (SIOH) study [ 14 ]. Both studies enrolled adults employed at Siriraj Hospital or residing in surrounding urban communities with the aim of investigating non-communicable diseases. Medication records were retrieved from the hospital EMR system as part of routine care and scheduled health checkups. All procedures adhered to the Declaration of Helsinki and were approved by the Institutional Review Board of the Faculty of Medicine Siriraj Hospital, Mahidol University (current study COA no. Si 235/2026). Cohort approvals included COA no. Si 647/2016 for the SIH cohort, and COA no. Si 381/2023 and Si 631/2019 for the SIOH cohort. Written informed consent was obtained from all participants at enrollment in the original studies. Genotype data from the original cohort datasets were accessed for research purposes on 4 June 2024. EMR data for this secondary analysis were accessed for research purposes on 1 April 2026, following Institutional Review Board approval of the current study. Genotyping was performed using the Infinium Asian Screening Array (ASA) v1.0 (Illumina, USA). Genotype quality control (QC) followed our previous work with minor modifications [ 14 , 15 ] and was conducted using PLINK v1.9 [ 16 ]. Participants with sex discrepancies were excluded, duplicated variants were removed, variants with call rate 0.086) was excluded to minimize bias in population frequency estimates. Hardy–Weinberg equilibrium filtering was not applied to PGx panel variants to avoid removing phenotype-defining markers. Variant coordinates were left-aligned and normalized to GRCh38 using bcftools v1.20 [ 18 ]. No genotype imputation was performed. The QC-passed, GRCh38-aligned dataset was used for downstream pharmacogenomic calling. Pharmacogenomic panel We constructed a pre-specified pharmacogenomic panel comprising 11 clinically relevant genes covered by the ASA v1.0: CYP2C19 [ 19 ], CYP2C9 [ 20 – 22 ] , CYP3A5 [ 23 ] , SLCO1B1 [ 20 ], ABCG2 [ 20 ] , VKORC1 [ 22 ], CYP4F2 [ 22 ] , TPMT [ 24 ] , NUDT15 [ 24 ] , UGT1A1 [ 25 ] and CYP2B6 [ 26 ] ( S1 Table ). Genes were selected based on the presence of CPIC-referenced markers on the array and the availability of CPIC guidelines with evidence level A or B for at least one medication used in the hospital setting. Genes requiring complex structural resolution (e.g., CYP2D6 ) were not included in the primary array-based panel. For each gene, we curated a minimal set of array markers from the ASA v1.0 manifest and mapped rsIDs to star-allele and functional definitions using CPIC and PharmVar resources. Markers were categorized as “required” or “optional” for star-allele determination. Required markers were necessary to distinguish CPIC-relevant star alleles and support diplotype assignment, whereas optional markers refined allele subtypes without changing CPIC phenotype category ( S1 Table ). Diplotype and phenotype calling Diplotype and phenotype calling were performed on the post-QC, GRCh38-aligned genotype dataset using the pharmacogenomic panel definitions above. For each gene, genotypes at panel markers were extracted from the VCF and interpreted in a reference/alternate-aware manner. Within each gene, genotype patterns at required markers (and optional markers when available) were matched to a gene-specific star-allele definition table ( S1 Table ). External statistical phasing was not performed. Allele definitions were implemented in a phase-insensitive manner when cis/trans configuration did not alter functional interpretation under CPIC phenotype categories. For TPMT, only the common decreased-function alleles *2, *3B and *3C were considered; genotype patterns were mapped to TPMT metabolizer status without distinguishing configurations such as *3A (cis combination of *3B and *3C) [ 24 ]. For UGT1A1 , rs887829 (*80) was used as a tag for the *28 haplotype [ 25 , 27 ]. For CYP2B6 , the *6 allele was approximated using the available markers rs3745274 and rs2279343 [ 26 ] ( S1 and S2 Tables ). These proxy definitions may not capture rarer alleles not represented on the array. A diplotype was assigned for each gene when a unique compatible combination could be identified. Diplotypes were translated into clinical phenotypes using gene-specific diplotype-to-phenotype mapping rules aligned to CPIC categories ( S2 Tables ). At the gene level, we applied a hybrid required/optional calling policy: a gene was considered called when all required markers were genotyped and consistent with panel definitions; if required markers were present but one or more optional markers were missing, a phenotype was assigned and flagged. If any required marker was missing or the genotype pattern did not match defined allele combinations, the gene was labeled undetermined and no phenotype was assigned. Callability (called/limited/undetermined) was summarized for each gene and used to define N_called (called + limited) denominators. Electronic medical record data and CPIC actionability EMR-derived medication exposures were ascertained from 1 January 2014–31 December 2024. Medication records were extracted for a pre-defined list of study drugs across all available strengths and formulations prescribed and/or dispensed by the hospital. A user was defined as having ≥1 prescription and/or dispensation record for a given drug during the observation period. Drug names were normalized to generic ingredients, and fixed-dose combinations were decomposed into constituent ingredients while retaining original product identifiers for traceability ( S3 Table ). For primary analyses, repeated records for the same individual and medication were collapsed into a participant-level ever-use indicator (0/1) for each medication. Statistical analysis All analyses were descriptive. For each pharmacogene, phenotype frequencies were calculated among callable participants (N_called) and reported as proportions with Wilson 95% confidence intervals. Gene-level callability (called, limited, undetermined) was summarized using counts and percentages. Participant-level burden was summarized as the number of CPIC-actionable genes per person and reported using counts and percentages, together with the median and interquartile range. For EMR-linked analyses, medication exposure was summarized as the number of users per medication (≥1 prescription/dispensation record from 2014–2024). Realized pharmacogenomic actionability was quantified for each CPIC level A/B drug–gene pair as the proportion of users carrying an actionable phenotype for the mapped gene (N_actionable_users/N_users). Individuals could contribute to multiple drug–gene pairs if they received multiple medications; drug-specific proportions were calculated independently within each medication’s user group and were not aggregated across drugs. For warfarin, VKORC1 rs9923231 and CYP4F2 rs2108622 genotype distributions and CYP2C9 phenotypes were summarized descriptively among warfarin users due to small sample size. Analyses were conducted in R v4.4.2. Visualizations were generated using ggplot2 and finalized in Adobe Illustrator v29.1 (Adobe, USA) Results Study cohort and pharmacogenomic panel callability A total of 5,031 participants from the SIH and SIOH studies with ASA genotyping data were available. After genotype QC (removal of duplicate variants, sex-discrepant participants, high-missingness variants and participants, and related individuals), 4,662 participants and 614,205 variants remained for analysis. A pre-specified pharmacogenomic panel comprising 26 variants across 11 genes was extracted from the QC-passed genotype dataset for downstream diplotype and phenotype calling ( Fig 1 ). Download: PNG larger image TIFF original image Fig 1. Study workflow for array-based pharmacogenomic calling and EMR linkage. Individuals from the Siriraj Health (SIH) and Siriraj OneHealth (SIOH) studies were genotyped using the Infinium Asian Screening Array (ASA) v1.0 (Illumina, USA). After genotype quality control (removal of duplicate variants, sex-discrepant samples, high-missingness variants/samples, and related individuals), 4,662 individuals with QC-passed genotypes were retained. A pre-specified pharmacogenomic panel (11 genes; 26 variants) was curated from CPIC level A/B gene-drug relationships and PharmVar/CPIC star-allele definitions and aligned to GRCh38. Diplotypes and phenotypes were assigned using a hybrid required/optional marker policy with pre-defined handling of selected special cases ( TPMT *2/*3B/*3C; UGT1A1 *80 tag; CYP2B6 *6 approximation). The primary outcome was the population pharmacogenomic landscape, and the secondary outcome was EMR-linked pharmacogenomic actionability. https://doi.org/10.1371/journal.pone.0355201.g001 Using a hybrid required/optional marker policy, gene-level callability was high across most genes, with an overall callability of 98.62% across all gene-by-participant results ( Table 1 ). Call rates (N_called/4,662) were 100% for CYP4F2 , CYP3A5 , and UGT1A1 , and ≥99% for VKORC1, CYP2C9, SLCO1B1, ABCG2, TPMT , and CYP2B6 . Lower callability was observed for CYP2C19 (N_called = 4,475; 95.99%) and NUDT15 (N_called = 4,209; 90.28%), reflecting a higher proportion of undetermined results at phenotype-defining markers. Marker-level allele frequencies and missingness for each panel variant are provided in Supplementary S6 Table . Overall, the high callability of the 26 panel variants supports the feasibility of using the ASA array for population-scale pharmacogenomic phenotype assignment in this cohort. Download: PNG larger image TIFF original image Table 1. Pharmacogenomic panel callability and denominators for phenotype frequency analyses in the cohort (n = 4,662). https://doi.org/10.1371/journal.pone.0355201.t001 Distribution of pharmacogenomic phenotypes and warfarin-related loci Gene-level phenotype distributions were summarized among callable participants for each gene (N_called) using the pre-specified calling and mapping rules ( Fig 2 and Supplementary S4 Table ); corresponding diplotype frequencies are reported in Supplementary S5 Table . Normal function/metabolizer phenotypes represented the largest proportion of callable results across genes, with varying contributions from decreased/intermediate and poor categories. For CYP2C19 (N_called = 4,475), intermediate and normal metabolizer phenotypes were similarly frequent (44.2% and 43.3%, respectively), while 11.2% were poor metabolizers; rapid and ultrarapid metabolizers were uncommon (1.2% and 0.1%). CYP2C9 results were predominantly normal metabolizer (90.6%; N_called = 4,655), with 9.4% intermediate and 0.02% poor metabolizers. For CYP3A5 (N_called = 4,662), intermediate (45.9%) and poor metabolizer phenotypes (41.5%) were more common than normal metabolizers (12.6%) ( Fig 2 and Supplementary S4 Table ). Together, these distributions indicate that clinically relevant non-normal metabolizer phenotypes—particularly for CYP2C19 and CYP3A5 —are common in the cohort. Download: PNG larger image TIFF original image Fig 2. Distribution of pharmacogenomic phenotypes and key warfarin-related genotypes in the study cohort (n = 4,662). Stacked bars show the proportion of individuals in each metabolizer/function category for CYP2C19, CYP2C9, CYP3A5, SLCO1B1, ABCG2, TPMT, NUDT15, UGT1A1 , and CYP2B6 , calculated among individuals with an assigned call for that gene (N_called; called + limited). Two warfarin-related loci, VKORC1 (rs9923231) and CYP4F2 (rs2108622), are shown as genotype categories. “Undetermined” denotes that no diplotype could be assigned due to missing required markers or unmatched patterns, whereas “Indeterminate phenotype ( NUDT15 )” indicates that a diplotype was assigned but mapped to an indeterminate NUDT15 phenotype under the pre-specified rules. Colors reflect clinical directionality where applicable (reduced function/sensitivity in yellow/red, normal in green, increased function/sensitivity in blue/purple, and unknown/other in greyscale). https://doi.org/10.1371/journal.pone.0355201.g002 For transporter genes, SLCO1B1 (N_called = 4,660) showed 20.5% decreased function and 1.3% poor function
## Related Clinical Research

- [Pharmacogenomic landscape in Thailand: Array-based PGx profiling and EMR-linked medication exposure](https://medichelpline.com/clinical-feed/pubmed-42545959.md) (DOI: 10.1371/journal.pone.0355201)
- [Widening medication adherence gap: impact on modern therapeutics and system solutions](https://medichelpline.com/clinical-feed/plos-medicine-0-the-widening-medication-adherence-gap.md)
- [Atorvastatin-loaded frankincense SNEDDS thermoresponsive hydrogel improves wound healing in a rat](https://medichelpline.com/clinical-feed/pubmed-42542257.md) (DOI: 10.1016/j.ijpharm.2026.127264)
- [Atorvastatin Adverse Event Confirmation in Older Medicare Beneficiaries Using Sequential Target-Tr](https://medichelpline.com/clinical-feed/medrxiv-11-beyond-signal-detection-sequential-target-trial-emulations-to-confirm.md)
- [24 HOUR OMEPRAZOLE (omeprazole 20 mg) tablet, delayed release [Grocery Outlet, Inc.]](https://medichelpline.com/clinical-feed/dailymed-updates-34-24-hour-omeprazole-omeprazole-20-mg-tablet-delayed-release-grocery-outlet-inc-.md)

## Navigation
- [← Back to Pharmacology Feed](https://medichelpline.com/clinical-feed/pharmacology.md)
- [← All Clinical Specialties](https://medichelpline.com/clinical-feed.md)
## Medical & Regulatory Disclaimer

> [!CAUTION]
> MedicHelpline content is structured for research, educational, and professional discovery purposes. It does not constitute individual medical advice, clinical diagnosis, or treatment recommendations.
> Always verify dosing, contraindications, and regulatory alerts against official product labeling and primary regulatory sources before clinical decision-making.