Clinical Function Assignment of NAT2 Alleles by the Clinical Pharmacogenetics Implementation Consortium Pharmacogene Curation Expert Panel Authors: Bailey M. Tibben, David W. Hein, Michelle Whirl‐Carrillo, Andrea Gaedigk, Michael T. Eadon, Rodney F. Minchin, Sotiria Boukouvala, Adalberto Rezende Santos, Roseann S. Donnelly, José A. G. Agúndez, Kelly E. Caudle Published: 2026 ABSTRACT NAT2 encodes arylamine N‐acetyltransferase 2, a key enzyme in the phase II metabolism of arylamines and arylhydrazines. NAT2 is highly polymorphic, resulting in variable distributions of rapid and poor metabolizers across global populations. Here, we detail the process undertaken by the Clinical Pharmacogenetics Implementation Consortium (CPIC) NAT2 Pharmacogene Curation Expert Panel (PCEP) to assign clinical function to NAT2 star (*) alleles using CPIC's standard terminology. Given the observed impact of NAT2 genetic variability on drug response, CPIC convened the NAT2‐PCEP to standardize clinical allele function assignments. The NAT2‐PCEP is comprised of multidisciplinary and international members, including researchers, clinicians, and implementers with expertise in pharmacogenomics and NAT2 molecular biology. Extensive in vitro and clinical literature was curated from PubMed and other sources to assess NAT2 genotype‐to‐phenotype concordance as well as the biochemical function of NAT2 star alleles. The NAT2‐PCEP assigned allele clinical function using CPIC's standard terminology (increased, decreased, uncertain, and unknown function) to 59 star alleles cataloged by the Pharmacogene Variation Consortium (PharmVar). Two alleles, NAT2*1 and NAT2*4, were assigned increased function (historically known as rapid), 40 alleles were assigned decreased function (historically known as slow), 10 alleles were assigned uncertain function, and seven alleles were assigned unknown function. Rigorous evidence review and in‐depth PCEP discussion were crucial in determining these function assignments. The findings reported here underscore the importance of standardized allele functional terms and diplotype‐to‐phenotype assignments to further the clinical implementation of NAT2 pharmacogenetic test results. Study HighlightsWHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC? NAT2 encodes arylamine N‐acetyltransferase 2, an enzyme involved in phase II metabolism of drugs (e.g., hydralazine, isoniazid, dapsone) and carcinogens. NAT2 is highly polymorphic, leading to rapid, intermediate, and poor metabolizer phenotypes that significantly affect drug efficacy and toxicity. There is strong evidence linking NAT2 genotype to phenotype and drug response.WHAT QUESTION DID THIS STUDY ADDRESS?This study aimed to standardize the clinical function assignment of NAT2 star alleles using CPIC's consensus‐based framework to enable consistent interpretation of pharmacogenetic test results for clinical implementation.WHAT DOES THIS STUDY ADD TO OUR KNOWLEDGE?The study provides a rigorous process for assigning clinical function to 59 NAT2 star alleles through expert integration of clinical, experimental, and computational evidence. This framework enhances transparency and consistency in pharmacogenetic reporting and prescribing decisions.HOW MIGHT THIS CHANGE CLINICAL PHARMACOLOGY OR TRANSLATIONAL SCIENCE?This work advances precision medicine by standardizing NAT2 allele function assignment for use in NAT2 CPIC Guidelines. Laboratories can harmonize genetic test interpretation, improving interoperability across healthcare systems. Researchers gain a foundation for future CPIC guidelines and substrate‐specific recommendations. Study Highlights WHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC? NAT2 encodes arylamine N‐acetyltransferase 2, an enzyme involved in phase II metabolism of drugs (e.g., hydralazine, isoniazid, dapsone) and carcinogens. NAT2 is highly polymorphic, leading to rapid, intermediate, and poor metabolizer phenotypes that significantly affect drug efficacy and toxicity. There is strong evidence linking NAT2 genotype to phenotype and drug response. WHAT QUESTION DID THIS STUDY ADDRESS? This study aimed to standardize the clinical function assignment of NAT2 star alleles using CPIC's consensus‐based framework to enable consistent interpretation of pharmacogenetic test results for clinical implementation. WHAT DOES THIS STUDY ADD TO OUR KNOWLEDGE? The study provides a rigorous process for assigning clinical function to 59 NAT2 star alleles through expert integration of clinical, experimental, and computational evidence. This framework enhances transparency and consistency in pharmacogenetic reporting and prescribing decisions. HOW MIGHT THIS CHANGE CLINICAL PHARMACOLOGY OR TRANSLATIONAL SCIENCE? This work advances precision medicine by standardizing NAT2 allele function assignment for use in NAT2 CPIC Guidelines. Laboratories can harmonize genetic test interpretation, improving interoperability across healthcare systems. Researchers gain a foundation for future CPIC guidelines and substrate‐specific recommendations. The Clinical Pharmacogenetics Implementation Consortium (CPIC) develops gene‐drug clinical practice guidelines that provide evidence‐based recommendations for adjusting therapy based on an individual's genetic makeup. 1 Because CPIC recommendations rely on the predicted phenotype derived from an individual's genotype (reported as a diplotype when using star allele haplotypes), having a standardized approach for phenotype assignment is essential to guideline development. CPIC assigns each allele a “Clinical Allele Function” using standard terminology (increased, decreased, normal, no, uncertain, or unknown function). 2 Phenotype (i.e., rapid, intermediate, or poor metabolizer) is determined by the combined “clinical allele functions” of the two alleles (or haplotypes) present in a person's genotype. For example, a NAT2 intermediate metabolizer is defined as having one increased function allele and one decreased function allele (e.g., NAT2*1/*5). 3 (Figure 1 ) This report discusses the NAT2 clinical allele function assignment process by the CPIC‐NAT2 Pharmacogene Curation Expert Panel (PCEP) and summarizes the supporting literature describing NAT2 genetic variation used to assign clinical allele function. CPIC updates allele function assignments as new evidence emerges, during guideline updates, or in response to inquiries from the pharmacogenetics community. 4 The most current NAT2 Allele Functionality Table is available on ClinPGx. 5 Relationship between CPIC allele function assignments, phenotype translation, and prescribing recommendations. CPIC translates pharmacogene genotypes into clinical prescribing recommendations through a coordinated process involving the Pharmacogene Curation Expert Panel (PCEP) and the Clinical Practice Guideline Panel (CPGP). The PCEP evaluates evidence to assign clinical function to individual pharmacogene alleles, while the CPGP evaluates evidence linking genotype‐predicted phenotypes to drug response and develops phenotype‐specific prescribing recommendations. Genotype‐to‐phenotype translation is developed jointly by the PCEP and CPGP by combining the function assignments of the maternal and paternal alleles. Because phenotype classification directly influences clinical recommendations, allele function assignment and genotype‐to‐phenotype translation are iterative and interconnected processes. NAT2 examples are highlighted in red. Although no NAT2 alleles are currently assigned a no function designation, this category may be used in the future as additional evidence becomes available. For a complete list of allele function assignments and diplotype‐to‐phenotype translations see the ClinPGx website. NAT2 encodes arylamine N‐acetyltransferase 2, which is involved in the phase II metabolism of arylamines and arylhydrazines, including hydralazine, isoniazid, dapsone, and sulfamethazine 6 as well as several carcinogens 7 and endogenous aliphatic amines. 8 NAT2 (MIM 612182) is highly expressed in the liver, small intestine, and colon. 6 Depending on the substrate, NAT2‐mediated acetylation may activate the compound or metabolize it to an inactive entity. 6 NAT2 is highly polymorphic, leading to trimodal (sometimes bimodal, depending on phenotype classification) distributions of rapid, intermediate, and poor (also known as “slow”) metabolizers across global populations, with a high degree of concordance between NAT2 phenotype and genotype 9 (see Supplemental Materials for more discussion). Variability in NAT2 activity significantly impacts drug efficacy and toxicity. For example, NAT2 poor metabolizers face higher risks of isoniazid‐induced hepatotoxicity or hydralazine‐induced lupus, while rapid and intermediate metabolizers may experience subtherapeutic drug levels. NAT1, NAT2, and the nonfunctional pseudogene NATP are encoded on the positive strand of chromosome 8p22. The NAT1 and NAT2 coding sequences are 87.5% homologous. 10 Despite this, NAT1 and NAT2 exhibit distinct substrate affinities that impact their unique physiological roles. 11 Additionally, variation in acetylator phenotype is primarily attributed to NAT2 genetic variants. NAT2 encodes a canonical mRNA transcript of 1,285 base pairs (bp), which contains one noncoding exon and a second exon that harbors the entire open reading frame of 870 bp, which produces an ~30 kDa protein of 290 amino acids. Currently, the Pharmacogene Variation (PharmVar) Consortium 12 defines over 50 distinct star alleles for NAT2; additional details regarding genetic variation of NAT2 can be found in the PharmVar NAT2 GeneFocus. 13 METHODS NAT2 Pharmacogene curation expert panel The NAT2 Pharmacogene Curation Expert Panel (NAT2‐PCEP) comprises multidisciplinary and international members, including researchers, clinicians, and implementers with expertise in pharmacogenomics and molecular genetics and a track record of publication and/or expertise in NAT2 gene function. CPIC is guided by the National Academy of Medicine Standards for Developing Trustworthy Clinical Practice Guidelines to minimize conflicts of interest of its authors and staff. All experts received training prior to their participation on the panel, the details of which are described elsewhere. 4 Briefly, training covered: (a) CPIC's evidence‐grading framework (Table 1 ), (b) standardized terminology for allele function assignments (Table S1 ), and (c) conventions for translating diplotypes into phenotypes for clinical implementation. The NAT2‐PCEP reviewed and discussed the literature informing NAT2 clinical allele functional assignments in a series of teleconferences. Feedback from panel members was crucial, especially to assess the nuances of functional calls for rare alleles and/or alleles which have limited evidence. Relevant committee discussions are included throughout to provide full transparency regarding the process of NAT2 allele function assignments. Table 1Assignment of strength of evidenceStrength of evidenceEvidence criteriaExamplesa Definitive Effect of the allele on acetylation function has been repeatedly demonstrated in independent clinical studies and has been upheld over timeFunction of a single variant has been heavily studied for > 35 years and is well‐establishedNo convincing, adequately powered contradictory evidence NAT2*5, NAT2*6, NAT2*7, NAT2*14 Strong At least two independent clinical studies provide evidence for the allele's role in acetylation function, in addition to at least one of the following types of evidence: ○Case reports○ In vitro data○Supportive computational data No convincing, adequately powered contradictory evidence N/AModerate At least two of the following provide evidence of the allele's role in acetylation function: ○≥ 2 patient cases○ in vitro experimental support○≥ 1 clinical study No convincing, adequately powered contradictory evidence Specifications Alleles defined by two or more well‐established variants (i.e., assigned definitive evidence) OR alleles defined by a well‐established variant in combination with a variant that has concordant computational evidence were assigned function with a moderate strength of evidence. See Figure 3 . NAT2*15, NAT2*29, NAT2*33, NAT2*50 Limited Limited but supportive evidence describes allele's role in acetylation function: case report(s), in vitro data, or computational predictionsNo convincing, adequately powered contradictory evidence Specifications Alleles defined by a well‐established variant (i.e., assigned definitive evidence) in combination with a variant that has discordant computational evidence, were assigned function with a limited strength of evidence. See Figure 3 . NAT2*19, NAT2*47, NAT2*48, NAT2*73 Inadequate – Uncertain function Fewer than two patient cases; little/no convincing in vitro dataWeak, insufficient, or conflicting evidenceNot strong enough to inform prescribing actionability NAT2*10, NAT2*21, NAT2*27 No evidence – Unknown functionThere is no literature describing function NAT2*55, NAT2*60, NAT2*61, NAT2*71 Note: CPIC framework for assigning strength of evidence to alleles applied to NAT2 with additional specifications made by the CPIC NAT2‐PCEP. a For a full list of NAT2 alleles and assigned function, see the NAT2 allele functionality table on the ClinPGx website. Assignment of strength of evidence Effect of the allele on acetylation function has been repeatedly demonstrated in independent clinical studies and has been upheld over time Effect of the allele on acetylation function has been repeatedly demonstrated in independent clinical studies and has been upheld over time Function of a single variant has been heavily studied for > 35 years and is well‐established Function of a single variant has been heavily studied for > 35 years and is well‐established No convincing, adequately powered contradictory evidence No convincing, adequately powered contradictory evidence At least two independent clinical studies provide evidence for the allele's role in acetylation function, in addition to at least one of the following types of evidence: ○Case reports○ In vitro data○Supportive computational data At least two independent clinical studies provide evidence for the allele's role in acetylation function, in addition to at least one of the following types of evidence: ○Case reports○ In vitro data○Supportive computational data ○Case reports Case reports ○ In vitro data In vitro data ○Supportive computational data Supportive computational data No convincing, adequately powered contradictory evidence No convincing, adequately powered contradictory evidence At least two of the following provide evidence of the allele's role in acetylation function: ○≥ 2 patient cases○ in vitro experimental support○≥ 1 clinical study No convincing, adequately powered contradictory evidence At least two of the following provide evidence of the allele's role in acetylation function: ○≥ 2 patient cases○ in vitro experimental support○≥ 1 clinical study At least two of the following provide evidence of the allele's role in acetylation function: ○≥ 2 patient cases○ in vitro experimental support○≥ 1 clinical study ○≥ 2 patient cases ≥ 2 patient cases ○ in vitro experimental support in vitro experimental support ○≥ 1 clinical study ≥ 1 clinical study No convincing, adequately powered contradictory evidence No convincing, adequately powered contradictory evidence Specifications Alleles defined by two or more well‐established variants (i.e., assigned definitive evidence) OR alleles defined by a well‐established variant in combination with a variant that has concordant computational evidence were assigned function with a moderate strength of evidence. See Figure 3 . Limited but supportive evidence describes allele's role in acetylation function: case report(s), in vitro data, or computational predictionsNo convincing, adequately powered contradictory evidence Limited but supportive evidence describes allele's role in acetylation function: case report(s), in vitro data, or computational predictions Limited but supportive evidence describes allele's role in acetylation function: case report(s), in vitro data, or computational predictions No convincing, adequately powered contradictory evidence No convincing, adequately powered contradictory evidence Specifications Alleles defined by a well‐established variant (i.e., assigned definitive evidence) in combination with a variant that has discordant computational evidence, were assigned function with a limited strength of evidence. See Figure 3 . Fewer than two patient cases; little/no convincing in vitro data Fewer than two patient cases; little/no convincing in vitro data Weak, insufficient, or conflicting evidence Weak, insufficient, or conflicting evidence Not strong enough to inform prescribing actionability Not strong enough to inform prescribing actionability Note: CPIC framework for assigning strength of evidence to alleles applied to NAT2 with additional specifications made by the CPIC NAT2‐PCEP. For a full list of NAT2 alleles and assigned function, see the NAT2 allele functionality table on the ClinPGx website. Inclusion of alleles The PCEP discussed and agreed to only include NAT2 alleles for clinical function assignment that have recently been rigorously reviewed by PharmVar (www.pharmvar.org/gene/NAT2). Substantial changes and updates have been made to NAT2 nomenclature, which are detailed in the PharmVar GeneFocus on NAT2 13 and the PharmVar NAT2 Read Me document. Alleles were included if cataloged in PharmVar version 6.2.15 (July 2025) after formal curation by the PharmVar NAT2 expert panel. Additional information, including a discussion of legacy NAT2 allele designations, is provided in the Supplemental Materials . Evidence review Evidence for assignment of allele clinical function was compiled via an extensive literature review, as described elsewhere. 4 Specifically, evidence for NAT2 was first compiled from a published review evaluating relationships between NAT2 genotype and phenotype. 9 In addition to the articles identified in this publication, literature searches were conducted with the search terms “NAT2 polymorphism and [substrate]”. Iterative searches were performed for caffeine, isoniazid, hydralazine, sulfamethazine, and sulfasalazine using PubMed. These substrates were selected because they represent the most widely studied probe drugs for NAT2 activity and provide the strongest and most consistent genotype–phenotype evidence, with the exception of caffeine (discussed in the supplemental materials). Additional articles were identified in NAT2 meta‐analyses and by citation tracing. Exclusion criteria were articles not in English, reviews, a lack of genotype data, the absence of functional studies, and/or the lack of direct genotype‐to‐phenotype comparisons (Figure S2 ). For a complete list of evaluated publications, see the “evidence” tab of Supplementary File S2 . In addition to published literature, in silico variant effect prediction (VEP) tools were utilized to further support the functional assignments for each star allele. CPIC uses 10 separate tools that have previously been evaluated for their performance in predicting the impact of genetic variation on pharmacogene function, 4 , 14 including PolyPhen‐2, 15 AlphaMissense, 16 PROVEAN, 17 MutationAssessor, 18 VEST, 19 CADD, 20 SIFT, 21 Likelihood Ratio Test (LRT), 22 REVEL, 23 and PhD‐SNPg. 24 Scores from PROVEAN, MutationAssessor, VEST, SIFT, REVEL, and LRT were obtained from a centralized location using the OpenCRAVAT webserver. 25 Scores and predictions (benign/tolerated, damaging/pathogenic) were recorded in the “computational evidence” tab of Supplementary File S2 . Predictions were interpreted as follows: Damaging: ≥ 6 tools predict deleteriousness; Conflicting: 4–5 tools predict deleteriousness; and Neutral: ≤ 3 tools predict deleteriousness. VEP data alone never overruled experimental evidence but aided in the resolution of ties for alleles with limited data (e.g., NAT2*10). Of note, not all VEPs are capable of scoring variants that introduce a premature stop codon, thus c.589C>T (p.R197X) has data from four tools, rather than 10. Computational predictions were used only as supportive evidence and did not independently determine allele function; in all cases where computational data contributed to evidence strength, functional assignment was primarily driven by well‐characterized variants with established experimental or clinical evidence. Limitations of current in silico tools in pharmacogenomics should be considered when interpreting these classifications. Assigning Allele function Function and strength of evidence determination The CPIC NAT2‐PCEP adhered to CPIC's standardized framework for assigning allele clinical function to NAT2 star alleles defined by PharmVar. 4 Allele clinical function was assigned to describe the clinically relevant function of each allele and to distinguish clinically actionable from non‐actionable alleles. An allele is clinically actionable if, when present in the right gene dosage (e.g., homozygosity or in compound heterozygosity with another actionable variant), prescribing decisions will be altered from normal (i.e., standard starting dose) prescribing actions per the CPIC guideline. 5 (Figure 1 ) Evidence was graded using a tiered system with scores ranging from “no evidence” to “definitive” evidence (Table 1 ). Alleles with “no evidence” or “inadequate evidence” to support clinically relevant functional consequences were assigned unknown function or uncertain function, respectively, and are considered non‐actionable. All other alleles were assigned a clinical function with evidence levels from “limited” to “definitive” (Table 1 ) and are considered by CPIC to be actionable. CPIC employs standardized terminology for assigning both allele function (Table S1 ) and phenotype across pharmacogenes (Table 2 ). These terms were previously established through the Delphi method to achieve consensus among pharmacogenetic experts. 2 However, due to the frequently observed bimodal or trimodal distribution of the acetylator phenotype, the scientific and clinical NAT2 communities have a long history of dividing NAT2 phenotypes as follows: either (1) two groups of rapid (fast) or slow (poor) acetylators (in keeping with the slow acetylator phenotype being an autosomal recessive trait), or (2) three groups of rapid (fast), intermediate, or slow (poor) acetylators (in keeping with acetylation being an autosomal co‐dominant trait). The term “normal acetylator” has rarely been used. Likewise, individual NAT2 alleles have been termed “rapid” or “slow.” To maintain alignment with legacy terminology while adhering to CPIC standards, terminology modifications were made. Within CPIC guidelines, NAT2 alleles are categorized as either increased function or decreased function, while the legacy terms rapid and slow are provided in the “Allele Biochemical Functional Status” column of the NAT2 Allele Functionality Table. Phenotypes are classified as rapid metabolizer, intermediate metabolizer, or poor metabolizer, with the term “normal metabolizer” intentionally omitted due to the clinical relevance of all phenotype groups for prescribing decisions. 5 The use of the terms increased function and decreased function aligns with historical nomenclature while minimizing potential confusion. Activity encoded by the NAT2*1 reference allele is referred to as “increased function” rather than “normal function” to avoid discrepancies with legacy terms. NAT2*1 and NAT2*4 represent the highest activity alleles identified after decades of NAT2 research and thus are classified as “increased function” in accordance with CPIC terminology. We acknowledge that this designation does not preclude the future identification of alleles with even greater activity. Should such variants be identified, refinement of the allele function classification framework and CPIC standardized allele function terms may be warranted. Reaching consensus on terminology required extensive discussions among the NAT2 guideline committee, NAT2‐PCEP, and broader CPIC membership. While most experts were in support of implementing CPIC standardized allele function and phenotype terms, the decision was not unanimous (Figure S1 ). Table 2Assignment of predicted NAT2 phenotype based on genotypeMetabolizer phenotypeAcetylator statusGenotypeExample Genotypes For a full list of diplotypes, see the NAT2 Diplotype‐Phenotype Table on the ClinPGx website NAT2 rapid metabolizerFast acetylatorTwo increased function alleles *1/*1, *1/*4, *4/*4 NAT2 intermediate metabolizerIntermediate acetylatorOne increased function and one decreased function allele *1/*5, *4/*6, *1/*7, *1/*14 NAT2 poor metabolizerSlow acetylatorTwo decreased function alleles *5/*5, *5/*6, *7/*7, *6/*14 NAT2 indeterminateN/Aa One or two alleles with uncertain or unknown function *10/*27, *5/*51, *7/*44 Note: Genotype‐to‐phenotype mapping conventions for NAT2. Note that the genotypes listed represent only a subset of all possible allele combinations (diplotypes). For a complete list of diplotypes see the ClinPGx website. Adapted from the CPIC® Guideline for Hydralazine and NAT2. 5 a N/A, phenotype not assigned. Assignment of predicted NAT2 phenotype based on genotype Note: Genotype‐to‐phenotype mapping conventions for NAT2. Note that the genotypes listed represent only a subset of all possible allele combinations (diplotypes). For a complete list of diplotypes see the ClinPGx website. Adapted from the CPIC® Guideline for Hydralazine and NAT2. 5 N/A, phenotype not assigned. RESULTS Evaluation of evidence Using the search criteria described above, 414 publications were identified, of which 116 met criteria and were included (Figure S2 , Supplementary File S2 ). Publications were evaluated using established criteria. 26 Summaries of findings for each allele can be found in the NAT2 Allele Functionality Table . Additional details related to evidence evaluation are provided in the Supplemental Materials . Assignment of Allele clinical function and strength of evidence The CPIC NAT2‐PCEP assigned an allele clinical function to 59 star alleles defined by PharmVar: two were assigned increased function (NAT2*1 and NAT2*4), 40 were assigned decreased function, 10 were assigned uncertain function, and seven were assigned unknown function ( NAT2 Allele Functionality Table ). There are 18 known single‐nucleotide variants (SNVs) that contribute to decreased NAT2 function, each with varying levels of supporting evidence regarding their functional impact; several of these variants are present in two or more star alleles. Figure 2 provides an overview of the NAT2 star alleles and how they compare to one another in terms of their core SNVs. NAT2 core allele comparison. NAT2 Allele Comparison. Each column represents an individual star allele. Each row represents a single‐nucleotide variant referenced according to NM_000015.3. Variants that are known to decrease function are colored dark red. Variants which are predicted to impair function based on computational evidence are colored light red. Variants which do not alter function are colored dark blue. Variants with uncertain or unknown function are colored light blue. Adapted from PharmVar CAVE Compare View. ac.499G>A (p.E167K) was shown to decrease NAT2 activity in vitro; however, in silico predictions indicate that the variant is neutral for protein function. Forty star alleles were assigned decreased function, and all except NAT2*19 and NAT2*48 contain one or more of the four core variants: c.191G>A (p.R64Q, rs1801279); c.341T>C (p.I114T, rs1801280); c.590G>A (p.R197Q, rs1799930); and c.857G>A (p.G286E, rs1799931), that have been extensively studied over the last 35+ years. Specifically, these variants have been identified in individuals who exhibit decreased NAT2 activity toward caffeine, 27 isoniazid, 28 , 29 and sulphonamides. 30 Thus, star alleles containing at least one of these well‐characterized variants were assigned decreased function, and all are considered clinically actionable, though the strength of evidence depends on which additional variants were or were not present (Figure 3 ). Strength of evidence assigned to alleles defined by multiple variants. Flow chart for assigning strength of evidence to NAT2 alleles defined by multiple genetic variants. “Known” function refers to variants defining NAT2 alleles that have been assigned a clinical function. “Unknown” function refers to variants defining alleles that have been assigned uncertain or unknown function according to the CPIC framework. Several rare star alleles were assigned a functional classification based on strong in vitro and in vivo evidence supporting the functional impact of a shared SNV found in multiple star alleles. While it is important to note that the full extent of the functional consequences of star alleles harboring multiple SNVs remains unknown, the potential range of decreased function does not appear to have clinical consequences. 5 Even so, to be completely transparent regarding the limited clinical evidence for rare star alleles containing an SNV of known functional impact and another SNV for which function is unknown, the NAT2‐PCEP developed specific rules for assigning functional classification and level of evidence to these alleles (Table 1 and Figure 3 ). First, star alleles that are defined by two or more SNVs, each with well‐established function, such as NAT2*15, which contains c.191G>A (p.R64Q) and c.590G>A (p.R197Q), were assigned decreased function with a moderate strength of evidence. Second, alleles that are defined by one SNV with well‐established function and another SNV that has only computational evidence to support function were also assigned a moderate strength of evidence (e.g., NAT2*37 and NAT2*49). In these cases, the rationale was to rely on the known function of the well‐characterized SNV when assigning function (i.e., c.590G>A, p.R197Q for NAT2*37 and c.341T>C, p.I114T for NAT2*49). Lastly, star alleles harboring one SNV with well‐defined function and one SNV with uncertain or unknown function (e.g., NAT2*47) were assigned decreased function based on the well‐characterized variant. These cases, however, received a limited strength of evidence due to the uncertain impact of the additional variant, either computationally or experimentally. Although the evidence for a particular star allele may have been sparse, the committee agreed that the definitive level of evidence assigned to the well‐characterized decreased function variants was sufficient to also assign decreased function to alleles that have these variants in combination with other rare variants of uncertain or unknown function (Table 1 , Figure 3 ). This strategy allowed the NAT2‐PCEP to maximize clinical utility, since all alleles with decreased function are considered by CPIC to be clinically actionable, even with a “limited” strength of evidence assignment. While consistent with CPIC methodology, these extrapolation approaches for rare alleles rely in part on expert‐defined rules and therefore involve a greater degree of expert judgment in the absence of direct experimental or clinical evidence. Of the 59 NAT2 star alleles, only two were assigned increased function: NAT2*1 and NAT2*4. Prior to the migration of NAT2 nomenclature to PharmVar, NAT2*4 was considered the reference allele. The majority of functional in vitro studies have used the activity level of protein encoded by NAT2*4 as a baseline for functional comparisons in the past and may continue to use NAT2.4 protein as the reference moving forward. The only difference between NAT2*1 and NAT2*4 is that the latter contains c.803G>A (p.R268K), which does not impact activity. 31 This variant is found at a global frequency of 59%, is the majority allele at this position across all gnomAD genetic ancestry groups, and is found within many other NAT2 star alleles (Figure 2 ). Clinically, individuals homozygous for NAT2*4 or compound heterozygous for NAT2*1 and NAT2*4 exhibit rapid metabolism of caffeine, 27 sulfamethazine, 30 dapsone, 32 and isoniazid, 33 as compared with individuals hetero‐ or homozygous for decreased function alleles. Additionally, individuals who are hetero‐ or homozygous for NAT2*4 metabolize hydralazine rapidly. 34 , 35 Collectively, this evidence supports the assignment of increased function for NAT2*1 and NAT2*4. Of the remaining star alleles, 10 were assigned uncertain function, and seven were assigned unknown function based on either inadequate and/or conflicting data or the lack of literature describing function, respectively. These star alleles are defined by a single SNV or in combination with c.803G>A and have not been tested experimentally in vitro or in vivo. While computational evidence predicts decreased function for some of these star alleles, computational evidence alone is not sufficient to meet CPIC's criteria for limited evidence except in unequivocal cases of early stop codon or complete gene deletions. Future research is warranted to characterize the function of these under‐studied variants. DISCUSSION Expert considerations Several key PCEP panel discussions are worth highlighting. Star alleles containing the core SNV c.857G>A (p.G286E, found in NAT2*7, *40, *59, and *68) are notable, as there are data indicating substrate‐specific activity of the produced NAT2 allozyme. In vitro studies have shown that the protein product of NAT2*7 exhibits decreased N‐acetylation activity toward numerous substrates, 36 , 37 while having increased affinity for dapsone, sulfamethazine, and sulfapyridine and decreased affinity for its donor substrate, acetyl‐CoA when compared with the protein product of NAT2*4. 37 , 38 , 39 In silico modeling predicts that c.857G>A disrupts the size and shape of the active site 40 which may explain why the allozyme encoded by NAT2*7 exhibits decreased acetylation activity toward some substrates, but not others. However, NAT2*7 is not assigned a substrate‐specific function at this time, as evidence shows that activity encoded by NAT2*7 is decreased toward hydralazine and isoniazid, drugs that have been prioritized for CPIC guidelines. 28 , 34 , 35 , 41 Potential future guidelines for other substrates may need to consider substrate specificity 39 when evaluating clinical function and associated prescribing recommendations for alleles containing the c.857G>A SNV, highlighting the need for future studies to better define substrate‐specific effects and their clinical relevance. While no NAT2 star alleles are currently classified as having “no function”, NAT2*27, *32, and *48 were considered for this designation. NAT2*27 contains c.589C>T which introduces a premature stop codon (p.R197X) that leads to a truncated protein predicted to be nonfunctional. 42 Although it lacks about one‐third of the coding sequence, it remains unknown whether, or to which degree, the truncation impacts function. In silico predictions suggest decreased or no activity, but due to a lack of in vitro or in vivo evidence and uncertainty about the integrity of the active site, it was assigned uncertain function. NAT2*32, defined by c.203G>A (p.C68Y) and c.341T>C (p.I114T), was considered for no function based on p.C68Y, which affects a key active site residue that is integral for acetyl transfer. 43 While in vitro studies suggest substitutions at residue p.C68 abolish enzymatic activity, 43 the committee could not reach a consensus due to the lack of clinical data suggesting abolished activity and instead classified NAT2*32 as decreased function based on the well‐documented function of c.341T>C (p.I114T). Similarly, NAT2*48, containing c.364G>A (p.D122N, a catalytic residue), was considered for no function based on in vitro evidence of undetectable enzymatic activity and protein expression. 38 Although these findings suggest a potentially severe biochemical impact, due to the absence of in vivo data confirming complete loss of function the committee could not reach consensus on an assignment of “no function,” and the allele was therefore assigned decreased function. This conservative approach ensures consistency in NAT2 allele function assignment while acknowledging that the available biochemical data may indicate a more profound loss of activity. The distinction between decreased and no function remains relevant for gene function discussions, but clinically, individuals with two decreased function alleles, two no function alleles, or one of each are all classified as poor metabolizers. Of note, these phenotype mappings are currently applicable to hydralazine and may need to be re‐evaluated for other drugs where complete loss of function potentially increases risk. In silico VEPs were used to complement existing experimental data supporting allele function assignments (Supplementary File S2 ). Notably, 10 variants, c.51C>G (p.N17K), c.264G>A (p.M88I), c.403C>G (p.L135V), c.472A>C (p.I158L), c.491 T>C (p.I164T), c.518A>G (p.K173R), c.536A>G (p.H179R), c.613A>T (p.M205L), c.803G>A (p.R268K), and c.857G>A (p.G286E), were predicted to be neutral or tolerated by all 10 VEPs tested. Most of these variants have not yet been tested experimentally, and additional data are needed to determine whether they alter NAT2 function. The c.803G>A variant, defining NAT2*4, has consistently been shown to have no impact on N‐acetylation activity in vitro and in vivo. Interestingly, although computational tools predict c.857G>A to be tolerated, alleles containing this variant presented with decreased function experimentally, which may be attributed to substrate‐specific effects, as discussed above. This result highlights that in silico predictions are not always accurate. As such, alleles with computational data only are not capable of reaching even a limited strength of evidence, and no NAT2 star alleles were assigned a clinically actionable function solely based on VEP. No variants were unanimously predicted to impair protein function; however, c.203G>A (p.C68Y) and c.364G>A (p.D122N) were predicted to be deleterious by nine and eight of the ten tools, respectively. In both cases, the remaining predictor(s) provided an ambiguous result. As discussed above, both were considered for a “no function” classification due to undetectable in vitro N‐acetylation activity, but the committee opted for “decreased function” due to the absence of supporting clinical data. Additionally, c.589C>T (p.R197X), resulting in a premature stop codon, was consistently predicted to impair function by all VEPs capable of such assessments. In some cases, VEPs influenced final function assignments when experimental evidence was limited; for instance, NAT2*10 is predicted to be tolerated by nine of 10 tools. Although in vitro evidence suggests reduced enzymatic activity for this allele, 38 the available evidence is limited and not supported by clinical genotype–phenotype data, and was therefore considered insufficient to support a clinical functional classification. The conflicting nature of the available data led to its assignment as “uncertain function.” Overall, VEPs provide valuable supplementary information to strengthen available data. Genetic test interpretation and clinical relevance According to the Genetic Testing Registry, there are currently 11 laboratories worldwide reporting NAT2 genotype using 16 available tests. Tests are performed using targeted SNV detection, next‐generation or Sanger sequencing of the NAT2 coding region, or allele‐specific PCR. Although sequencing the NAT2 coding region allows identification of all NAT2 variants, the use of targeted SNV panels may be more cost effective and can still provide the necessary information for dose adjustments. 44 Although NAT2 does not yet have recommendations for clinical genotyping from the Association for Molecular Pathology, various SNV panels have been proposed for efficient prediction of NAT2 phenotype. One study measuring NAT2 acetylation activity in cryopreserved human hepatocytes found that a simplified 4‐SNV panel consisting of c.191G>A, c.341T>C, c.590G>A, and c.857G>A predicted NAT2 phenotype with an accuracy of 98.4%. 45 These four SNVs, along with the c.803G>A SNV ‐ which distinguishes NAT2*1 and NAT2*4 and is present on multiple NAT2 star alleles ‐ define the highest frequency alleles. c.191G>A, c.341T>C, c.590G>A, and c.857G>A have accurately predicted acetylator phenotype in the majority of individuals in diverse populations. 9 , 45 It is important to note that genotyping only for common NAT2 SNVs (e.g., using the above‐mentioned 4‐SNV panel) provides sufficient information to default to the common allele clusters (NAT2*1, *5, *6, *7, and *14), assuming that the high‐frequency c.803G>A variant is present on *6, *7, and *14. However, these SNVs are often compatible with more than one possible diplotype, and not all will translate to the same phenotype. For example, if testing identifies heterozygosity for c.341T>C and c.590G>A, the diplotype could default to NAT2*5/*6 predicting a poor metabolizer. However, c.341T>C is found in 12 other alleles, and c.590G>A is found in 18 other alleles, and both variants are present on NAT2*30, meaning the diplotype could be NAT2*1/*30 predicting an intermediate metabolizer and requiring a different CPIC recommendation for hydralazine dosing. 5 Furthermore, an increasing number of alleles have none of the common SNVs covered by a limited panel; such alleles would escape detection and be assigned as NAT2*1. It is therefore important to understand and appreciate these limitations as these may impact prescribing recommendations. Genotype‐to‐phenotype determination The assignment of phenotypes to the many possible genotypes for a given pharmacogene is a key step in the process of clinical implementation of pharmacogenetic test results. Specifically, the use of standardized terms 2 to distill complex genomic information into phenotypes facilitates interoperability of pharmacogenetic results and informs clinical decision‐making in the healthcare system. Genetic test results are reported as genotypes, or as diplotypes reflecting the combination of two star alleles, each representing a haplotype inherited from each parent (e.g., NAT2*1/*5), and phenotypes are assigned based on the reported diplotype and the corresponding clinical allele function assignments, as summarized in Table 2 . In the case of NAT2, individuals with two increased function alleles are categorized as rapid metabolizers. Individuals with one increased function allele and one decreased function allele are categorized as intermediate metabolizers. Finally, individuals with two decreased function alleles are categorized as poor metabolizers. Individuals with one or two uncertain or unknown function alleles are categorized as “indeterminate.” There are several NAT2 star alleles with varying levels of decreased function, and thus, different combinations of alleles create diplotypes with a range of decreased activity, according to in vitro functional assays. Specifically, NAT2*6/*6 gives rise to an “ultraslow” acetylation phenotype in vitro and, to a lesser extent, in patients. 27 , 39 , 46 , 47 However, the observed difference is more pronounced when activity is measured in in vitro systems, and currently there is not enough evidence supporting a clinically relevant difference to warrant alternate prescribing recommendations between NAT2*6/*6 patients and patients with other combinations of decreased function alleles. Historically, the NAT2 research community has inconsistently identified the intermediate metabolizer group, particularly when assessing clinical studies. Biochemically, NAT2 alleles represent an exquisite example of variant pharmacogene function. That is, experiments in cryopreserved human hepatocytes have shown that the expression of two increased function alleles leads to rapid acetylation activity, the expression of two decreased function alleles leads to slow acetylation activity, and the expression of one increased function and one decreased function allele leads to acetylation activity that falls precisely in between rapid and slow activity. 45 However, the intermediate metabolizer group is not always clearly defined in patient data. In some studies, there is a clear trimodal distribution, 28 , 34 while other studies show a bimodal distribution 48 , 49 or an overlap of intermediate and rapid metabolizers. 46 , 50 There are many factors which may contribute to the phenotypic distribution of NAT2 activity, such as the probe drug used, co‐administration of medications, acute illness, and other environmental factors. Compared with phenotyping, NAT2 genotype is more easily and inexpensively obtained, and ideally is available before the drug is prescribed. Thus, NAT2 genotyping represents an important method to predict NAT2 phenotype and adjust medication dosage accordingly. FUNDING This work was supported in part by ALSAC. PharmVar is supported by the Children's Mercy Research Institute (CMRI). CONFLICT OF INTEREST The authors declared no competing interests for this work. AUTHOR CONTRIBUTIONS B.M.T., D.W.H., K.E.C., R.S.D., M.W.C., A.G., and J.A.G.A. wrote the manuscript. K.E.C., R.S.D., M.W.C., A.G., and B.M.T. designed the research. B.M.T., D.W.H., M.W.C., A.G., M.T.E., R.F.M., S.B., A.R.S., and J.A.G.A. performed the research and analyzed the data. Supporting information Data S1. Data S2.