Understanding MASLD’s genetic architecture is essential for deciphering the disease’s clinical heterogeneity and individual risk of progression. Genome-wide association studies (GWAS) have consistently identified several variants showing robust associations with MASLD susceptibility and disease severity. Key loci include patatin-like phospholipase domain-containing 3 (PNPLA3), transmembrane 6 superfamily member 2 (TM6SF2), sorting and assembly machinery component 50 homolog (SAMM50), glucokinase regulator (GCKR), membrane-bound O-acyltransferase 7 (MBOAT7), and protective hydroxysteroid 17-beta dehydrogenase 13 (HSD17B13) variants. These genetic determinants can be broadly classified into hepatic fat-promoting risk alleles and protective variants, each modulating distinct lipid and inflammatory pathways^[21,22]. Complementing traditional GWAS approaches, whole-exome sequencing coupled with machine learning algorithms has enabled the discovery of rare variants associated with severe phenotypes^[21,22], thus establishing a comprehensive genomic foundation for precision risk stratification and targeted gene therapy development [Table 1].

The PNPLA3 rs738409 C>G (I148M) variant is the single most influential risk allele for MASLD, showing robust associations with hepatic fat content, lobular inflammation, hepatocellular ballooning, advanced fibrosis, and HCC across diverse populations^[24]. Mechanistically, the I148M substitution impairs triglyceride hydrolase activity, causing mutant PNPLA3 protein accumulation on lipid droplet surfaces and obstruction of normal lipolysis, thereby amplifying lipotoxic hepatocellular stress^[23]. Although these associations are robustly replicated, the reported effect estimates, particularly for HCC risk, vary substantially across cohorts, likely reflecting differences in alcohol co-exposure, ancestry, and underlying liver disease etiology^[25].

TM6SF2 rs58542926 C>T (E167K) is the second major locus, with carriers exhibiting elevated hepatic triglyceride content but reduced circulating low-density lipoprotein (LDL) and very-low-density lipoprotein (VLDL) cholesterol. These phenotypes suggest compromised efficiency of hepatic lipoprotein secretion^[26]. Clinically, this variant is associated with more severe steatosis and an elevated risk of advanced fibrosis and cirrhosis^[27,28]. When combined with PNPLA3 risk variants, TM6SF2 identifies a high-risk MASLD subgroup with a pronounced hepatic disease burden^[47,48].

Variants in GCKR and MBOAT7 modulate MASLD susceptibility by altering carbohydrate and phospholipid metabolism, respectively. GCKR rs1260326C>T (P446L) and related functional variants enhance glucokinase activity, thereby increasing hepatic glucose uptake and de novo lipogenesis (DNL), predisposing to steatosis and hypertriglyceridemia, particularly in insulin-resistant individuals^[30,31]. MBOAT7 rs641738 C>T reduces lysophosphatidylinositol acyltransferase expression, disrupting membrane phospholipid composition, promoting inflammatory signaling, and increasing vulnerability to steatohepatitis and fibrosis progression^[33,34]. SAMM50 rs738491 - encoding an outer mitochondrial membrane protein essential for mitochondrial morphology - modulates reactive oxygen species (ROS) levels and β-oxidation capacity^[36,37] and demonstrates a significant association with MASLD in the Korean GWAS^[38].

In contrast to risk alleles, HSD17B13 rs72613567:TA confers robust protection against progression from isolated steatosis to steatohepatitis, advanced fibrosis, and HCC^[41,43]. Carriers demonstrate lower necroinflammatory activity and milder fibrosis via attenuation of lipotoxic injury^[44]; mechanistically, HSD17B13 phosphorylation at S33 promotes hepatic fat catabolism, whereas the S33A mutation impairs this protective function, thereby promoting steatosis and MASH severity^[42].

Beyond HSD17B13, MARC1 (also known as MTARC1) has emerged as another genetically validated protective locus in MASLD. A common missense variant (rs2642438, p.A165T) in MARC1, first identified through GWAS for all-cause liver cirrhosis, confers protection against hepatic steatosis, cirrhosis, and HCC^[46]. The protective p.T165 allele results in reduced MARC1 protein stability and increased proteasomal degradation, leading to approximately 50% lower hepatic MARC1 protein levels^[45].

Individual risk variants can be aggregated into PRS, capturing cumulative genetic liability for advanced MASLD phenotypes^[49]. Simple PRS incorporating PNPLA3, TM6SF2, and HSD17B13 significantly improve prediction of significant fibrosis and histologic severity beyond clinical factors alone in population-based cohorts^[50,51]. More sophisticated partitioned PRS approaches, which dissect “liver-centric” vs. “cardiometabolic” genetic components, have recently identified distinct MASLD subtypes with divergent patterns of hepatic fibrosis, HCC risk trajectories, and systemic cardiometabolic complications, providing a molecular framework for precision endotyping^[52]. However, most current PRS validation studies are cross-sectional, and prospective evidence demonstrating the clinical utility of risk-based interventions is still emerging.

While GWAS identify risk loci, eQTL mapping provides the mechanistic link between genetic variants and their functional consequences in gene expression. Conventional eQTL approaches using healthy liver tissues have profiled thousands of gene-regulatory variants; however, they fail to capture dynamic regulatory effects relevant to specific disease contexts^[53,54].

Yoo et al. devised a disease-specific eQTL screening pipeline using liver biopsies and blood samples from 125 Korean individuals (83 with biopsy-proven MASLD and 42 without MASLD)^[53]. The authors identified 242 MASLD-specific eQTLs, which are genetic variants whose regulatory effects on gene expression are detectable only in the diseased liver environment, not under healthy conditions, through eQTL analyses of 21,272 transcripts and over 3.2 million genotyped/imputed SNPs. Among these, alanine-glyoxylate aminotransferase 2 (AGXT2) emerged as a top MASLD-eQTL: the eSNP rs2291702 constituted a significant cis-eQTL specifically in the MASLD group (P = 7.21 × 10^-9) but not in the no-MASLD controls (P = 0.38) or GTEx liver data (P = 0.31). Notably, AGXT2 downregulation was restricted to rs2291702:CC homozygotes within the MASLD group, demonstrating genotype-dependent gene regulation under disease conditions. This finding was replicated in an independent cohort of 162 Korean individuals (combined P = 8.05 × 10^-8). Functional validation confirmed that AGXT2 knockdown exacerbated MASLD diet-induced liver fibrosis by increasing endoplasmic reticulum (ER) stress-mediated hepatocellular death, whereas AGXT2 overexpression attenuated fibrosis and steatosis^[53]. These results established AGXT2 as a genotype-dependent therapeutic candidate and demonstrated that disease-specific eQTL mapping could uncover regulatory variants not detected by conventional approaches.

Building on this framework, Hong et al. advanced the approach to single-cell resolution by performing single-nucleus RNA sequencing (snRNA-seq) on liver biopsies from 25 patients with MASLD and 23 controls^[54]. This single-nucleus eQTL analysis, employing a Poisson mixed-effects model, identified over 3,500 liver-eQTLs across four major liver cell types [hepatocytes, cholangiocytes, hepatic stellate cells (HSCs), and endothelial cells], with the majority being cell type-specific (58.3%-72.8%). A critical innovation was the identification of cell-state-interacting eQTLs (ieQTLs), which are genetic regulatory effects whose magnitude varies depending on the cell’s functional state or zonal position within the hepatic lobule. Strikingly, ieQTLs demonstrated significant enrichment for MASLD trait heritability [enrichment odds ratio (OR) = 10.27], substantially exceeding that of conventional liver-eQTLs (OR = 5.39). Among the most notable ieQTLs, the protective variant rs28636836 in HSD17B13 exhibited zone-specific regulation with the strongest eQTL effect in periportal hepatocytes, while CYP4V2, a druggable fatty acid metabolism gene, showed periportal-specific regulation. These zonation-dependent eQTLs have immediate implications for patient stratification in RNA-based therapeutic trials and for the discovery of zone- and genotype-specific treatment targets.

The study further integrated transcription factor (TF) activity analysis with iQTLs, revealing 601 functional “quartets” comprising TFs, cell states, ieSNPs, and ieGenes. Based on this framework, the investigators pinpointed the FOXO1-rs13395911-EFHD1 axis: in maladapted hepatocytes (characterized by reduced Hep-M12 metabolic module expression), decreased FOXO1 activity led to the loss of EFHD1 expression specifically in rs13395911:TT genotype carriers^[54]. EFHD1 protected against lipotoxicity-induced ER stress, mitochondrial dysfunction, and apoptosis in cell lines and human hepatic organoids, suggesting that it is a genotype-dependent therapeutic target for MASLD.

Genetic risk factors for MASLD exhibit profound ethnic heterogeneity, necessitating localized genomic research to develop genomically and ethnicity-adapted clinical management strategies. The global distribution of key risk alleles varies markedly, directly influencing disease prevalence and severity across populations.

The PNPLA3 rs738409 C>G variant exemplifies striking ethnic disparities, with the risk G allele reaching approximately 50% frequency in Hispanics of Mexican, Central, or South American descent, correlating with their disproportionately high rates of MASH, advanced fibrosis, and HCC^[25]. Conversely, African and African-American populations exhibit the lowest G-allele frequency (AF) (approximately 12%-14%), conferring protection against hepatic steatosis despite elevated systemic insulin resistance and obesity burdens. Individuals of European ancestry occupy an intermediate frequency range (approximately 23%), showing moderate MASLD susceptibility.

TM6SF2 rs58542926 C>T contributes to ethnic variation: prevalence reaches 7.2%, 4.7%, and 3.4% in individuals of European ancestry, Hispanics, and African-Americans, respectively, with variant carriers exhibiting increased hepatic triglyceride retention due to impaired VLDL secretion^[29]. MBOAT7 rs641738 C>T shows the strongest associations with MASLD progression and fibrosis in European-Caucasian cohorts, while demonstrating inconsistent-to-absent associations in Asian and African-American populations^[35,55].

Meta-analysis has revealed that lean MASLD prevalence [defined as body mass index (BMI) < 25 kg/ m² in Caucasians or BMI < 23 kg/ m² in Asians] was 12% in Asia, 10.2% in the Middle East, and 9.2% in Western countries^[56]. In lean MASLD cohorts, genetic predisposition - particularly PNPLA3 and SAMM50 variants - predominates over systemic adiposity in driving MASLD, explaining MASLD occurrence in lean individuals independent of manifest metabolic syndrome^[39]. This ancestry-driven genetic architecture necessitates ethnicity-specific predictive models extending beyond traditional obesity paradigms.

Within Korean populations, the GENIE cohort study established significant associations between PNPLA3 and SAMM50 genes and both MASLD presence and severity^[40]. Additional GWAS identified associations of PNPLA3 with metabolic dysfunction-associated fatty liver disease, specifically with overweight/obese phenotypes^[57]. PNPLA3 variants are correlated with increased immune cell infiltration and advanced fibrosis progression^[58]. Notably, a disease-specific eQTL approach applied to the Korean biopsy-proven MASLD cohort revealed population-specific regulatory dynamics: the AGXT2 eSNP rs2291702 showed variable AFs across populations, with the reference T allele predominating in African populations (mean AF = 0.802) but being minor in East Asians and Koreans (AF = 0.327 and 0.330, respectively), implying differential susceptibility to the AGXT2-dependent MASLD pathway across ethnicities^[53]. These Korean-specific genetic discoveries collectively support the development of ancestry-tailored PRS and precision screening strategies targeting the evolving MASLD burden in East Asian populations.

Although the Korean cohort findings summarized above provide robust East Asian-specific evidence, they should be interpreted within the broader global landscape of MASLD genetic susceptibility. The disproportionate burden of MASH, advanced fibrosis, and HCC in Hispanic populations is largely attributable to the high frequency of the PNPLA3 I148M risk allele (approximately 50%)^[25], whereas individuals of African ancestry exhibit relative protection from hepatic steatosis despite the high prevalence of obesity and insulin resistance, partly owing to the low frequency of the PNPLA3 risk allele^[25,29]. South Asian, Middle Eastern, and African populations remain markedly underrepresented in MASLD genomic studies despite the rapidly rising disease burden^[5]. Consequently, ancestry-diverse genomic studies and population-specific validation of risk variants and PRS are essential to ensure that emerging precision medicine approaches are equally applicable across global populations, rather than disproportionately benefiting individuals of European ancestry.

Epigenetic regulation - encompassing DNA methylation, histone modifications, and non-coding RNAs - bridges environmental exposures (diet, obesity, insulin resistance, dysbiosis) with heritable alterations in gene expression driving MASLD initiation, progression, and potential reversal. DNA methylation patterns dynamically track disease progression from steatosis through fibrosis, with epigenome-wide association studies (EWAS) identifying differentially methylated regions that are strongly correlated with fatty liver index and capture longitudinal changes in steatosis burden^[59].

Korean liver biopsy EWAS (n = 106) identified hypermethylation and downregulation of complement genes (C1R, C1S, C3, C6, C4BPA, SERPING1) alongside hypomethylation and upregulation of C5AR1, C7, and CD59, patterns that correlate with MASLD histological severity (steatosis, inflammation, fibrosis). These findings were validated in public datasets (GSE180474) and MASH mouse models^[60]. Recent whole-genome methylation sequencing in MASLD cohorts identified sex-specific DNA methylation signatures and circulating cell-free DNA biomarkers (e.g., TM4SF5), facilitating highly accurate non-invasive fibrosis stage prediction^[61]. These observations underscore DNA methylation as a pivotal epigenetic marker that bridges genetic susceptibility with environmental factors, providing a framework for precision medicine through non-invasive diagnostics and personalized longitudinal monitoring across the MASLD spectrum.

Beyond descriptive methylome profiling, recent mechanistic studies have directly implicated DNA methyltransferase 1 (DNMT1) as a central epigenetic driver of MASLD progression. Sohn et al. conducted integrated analyses of the hepatic transcriptome (n = 131) and DNA methylome (n = 106) from patients with biopsy-confirmed MASLD and demonstrated consistent upregulation of DNMT1 expression across multiple independent MASLD cohorts, correlated positively with histological severity, including hepatocellular ballooning, lobular inflammation, NAFLD activity score (NAS), and fibrosis stage^[62]. DNMT1 expression was positively correlated with insulin resistance parameters (homeostatic model assessment for insulin resistance, circulating insulin, HbA1c), linking epigenetic dysregulation to metabolic perturbation. A meta-analysis integrating multiple cohorts confirmed reproducible DNMT1 upregulation, whereas DNMT3A and DNMT3B did not show consistent changes^[62]. Although the reproducibility of DNMT1 upregulation across multiple cohorts is relatively well-established, several other epigenetic observations summarized in this section, including specific differentially methylated regions and chromatin accessibility signatures, are based on single-cohort data and require independent validation.

DNA methylome profiling in both human MASLD samples and a diet-induced mouse model revealed significant enrichment of binding motifs of key hepatic TFs, particularly hepatocyte nuclear factor 4α (HNF4α) and peroxisome proliferator-activated receptor α (PPARα), among MASLD-associated hypermethylated differentially methylated positions^[62]. Importantly, the expression of HNF4α and PPARα was inversely correlated with histological indicators of disease activity (ballooning, lobular inflammation) and with DNMT1 expression in human MASLD samples, indicating that DNMT1-mediated hypermethylation functionally suppressed these master metabolic regulators during disease progression.

Histone modifications, such as acetylation and methylation, are important epigenetic processes controlling gene expression in chronic diseases, such as MASLD fibrosis. These reversible histone marks dynamically regulate profibrogenic gene expression in HSCs, the primary mediators of extracellular matrix deposition and fibrosis progression^[63].

Complementing studies on histone modification, genome-wide chromatin accessibility profiling using the assay for transposase-accessible chromatin with sequencing (ATAC-seq) has provided a comprehensive epigenomic reference for MASLD progression. Kang et al. performed ATAC-seq on 44 liver biopsy samples spanning the full histological spectrum of MASLD - from healthy controls through isolated steatosis to fibrotic MASH - in a Korean biopsy-proven cohort^[64]. A total of 112,303 merged accessible chromatin regions were identified. Principal component analysis and correlation heatmaps clearly distinguished disease stages by their chromatin accessibility profiles.

Compared with isolated steatosis, fibrotic MASH showed 272 up-regulated and 1,137 down-regulated differentially accessible regions (DARs) (false discovery rate < 0.05)^[64]. Functional enrichment analysis of down-regulated DARs in fibrotic MASH revealed significant associations with hormone response, cell cycle regulation, steroid metabolic processes, and liver development pathways. TF binding motif enrichment and de novo motif analyses within these DARs identified candidate biomarkers, including ETS family members (ETV1, ETV4, ELF3, GABPA), EWSR1-FLI1, IKZF1, SPI1, STAT2, and ZNF384, which could serve as epigenomic diagnostic markers distinguishing MASLD progression stages^[64]. These chromatin accessibility signatures complement DNA methylation and histone modification data, collectively delineating the regulatory specificity of non-coding genomic regions affected by genetic and epigenetic alterations during MASLD progression.

MicroRNAs (miRNAs) orchestrate MASLD pathogenesis through post-transcriptional modulation of lipid homeostasis, inflammatory signaling, and extracellular matrix remodeling. Liver-enriched miR-122, comprising approximately 70% of total hepatic miRNAs, regulates cholesterol biosynthesis and fatty acid metabolism. Circulating miR-122 levels increase in steatosis and MASH due to hepatocyte damage, functioning as a sensitive biomarker^[65,66]. miR-192, extensively studied alongside miR-122, is upregulated in MASH relative to isolated steatosis, contributing to disease progression through inflammation modulation (e.g., M1 macrophage polarization via Rictor/Akt/FoxO1 signaling)^[67]. Circulating miRNA panels provide non-invasive diagnostic, staging, and monitoring capabilities for MASLD/MASH, surpassing single-marker performance through multi-miRNA combinations^[68,69]. Therapeutic targeting of miRNAs, particularly miR-122 via antisense inhibition, reduces hepatic lipid accumulation, inflammation, and fibrosis in MASLD models, with demonstrated clinical translation in hepatitis C virus trials and emerging MASLD applications^[70].

Gene-environment interactions are key to understanding heterogeneous MASLD phenotypes and variable progression rates among individuals with similar metabolic risk factors. Disease expression reflects a dynamic interplay between inherited susceptibility alleles and modifiable environmental exposures, including systemic metabolic status, dietary patterns, adiposity distribution, and physical activity levels^[71,72].

Hepatic and peripheral insulin resistance represent major upstream drivers of MASLD. Insulin resistance amplifies adipose tissue lipolysis, delivering excess free fatty acids to the liver and stimulating DNL, thereby promoting intrahepatic triglyceride accumulation and lipotoxic injury^[73]. High-carbohydrate diets rich in fructose and refined sugars enhance DNL by upregulating ChREBP and SREBP1c TFs, while simultaneously promoting oxidative stress and gut-derived endotoxemia that aggravate steatohepatitis^[74,75]. Saturated fat-rich diets increase hepatocellular ceramides and diacylglycerols, impairing insulin signaling and amplifying inflammatory and fibrogenic responses^[76,77]. Obesity is a potent pro-inflammatory amplifier owing to adipokine secretion (elevated leptin and resistin; reduced adiponectin), which creates an inflammatory milieu that accelerates the transition from isolated steatosis to MASH and fibrosis^[78]. Sarcopenic obesity, whose prevalence in aging populations and Asian cohorts is increasing, compounds the risk of MASLD/MASH progression through reduced skeletal muscle glucose disposal capacity and altered myokine secretion. This worsens systemic insulin resistance and inflammation^[79,80]. Physical inactivity aggravates these pathways, whereas higher moderate-to-vigorous physical activity levels are associated with lower steatosis and fibrosis risk, independent of genetic and metabolic risk factors^[81,82].

The magnitude of MASLD induced by metabolic and dietary stressors is substantially modified by individual genotype. PNPLA3 rs738409 G (I148M) allele carriers exhibit more severe hepatic steatosis than non-carriers for equivalent adiposity levels, with genetic effects amplified by sugar-sweetened beverage consumption^[83]. PNPLA3 G-allele carriers experiencing ≥ 10 kg weight gain after early adulthood demonstrate markedly elevated MASLD prevalence even at non-obese BMI, illustrating the combined impact of genetic susceptibility and modest weight gain on disease expression^[84].

Similarly, TM6SF2 rs58542926 T-allele carriers exhibit reduced hepatic VLDL-triglyceride export, causing caloric and fat excess to preferentially manifest as intrahepatic triglyceride and cholesterol retention with amplified fibrosis risk despite relatively favorable circulating lipid profiles^[28,85]. MBOAT7 risk alleles alter hepatocellular phosphatidylinositol acyl-chain composition, enhancing pro-inflammatory signaling and HSC activation, thereby increasing fibrosis susceptibility for equivalent obesity and insulin resistance burdens^[86,87].

Gene-environment interactions extend to variable treatment responses to lifestyle modification. While weight reduction, Mediterranean-style dietary patterns, and lower-carbohydrate regimens remain foundational strategies for most patients, genetic profiling helps explain inter-individual variation in intervention efficacy^[72]. Depending on inherited polygenic risk, some individuals benefit from modest dietary and physical activity improvements, whereas others who particularly do not respond to metabolic targeted therapies [e.g., thyroid hormone receptor beta (THR-β) agonists and glucagon-like peptide-1 (GLP-1) receptor agonists] require more intensive integrated programs combining tailored nutrition, structured exercise, and emerging gene- and epigenome-targeted therapies for truly personalized MASLD management.

Gene-based therapy encompasses a diverse spectrum of strategies tailored to specific therapeutic objectives, ranging from transient gene modulation to permanent genomic correction. Transient approaches such as small-interference RNA (siRNA) and antisense oligonucleotides (ASOs) provide reversible control over malfunctioning genes^[92]. These evolving technologies - including high-precision base and prime editing - are currently being leveraged to restore metabolic homeostasis in liver diseases, with several clinical trials evaluating the suppression of key genetic targets to treat MASLD underway.

Epigenetic therapies aim to reprogram pathologic transcriptional states by targeting core mechanisms, including DNA methylation, histone modifications, and non-coding RNA-mediated gene regulation. In MASLD, accumulating evidence indicates that hepatocytes and HSCs acquire characteristic epigenetic alterations, including hypermethylation of lipid and cholesterol metabolism, more permissive chromatin at inflammatory and fibrogenic loci, and profoundly dysregulated miRNA networks^[99]. Given the inherent reversibility of epigenetic marks, epigenetic drugs provide opportunities to “reprogram” hepatic transcriptional states and function without altering underlying DNA sequences.

PRS represents key tools that operationalize genomic discoveries for MASLD risk stratification and individualized clinical care. PRS aggregates common variant effects in genes regulating hepatic lipid levels (PNPLA3, TM6SF2, MBOAT7, GCKR, HSD17B13, MARC1) into a quantitative measure capturing inherited disease predisposition. De Vincentis et al. demonstrated that hepatic fat PRS adds prognostic information beyond traditional fibrosis scores (FIB-4), providing a more granular assessment of progression to severe liver disease and unmasking high-risk individuals within the clinically ambiguous intermediate-risk category who may benefit from intensified surveillance^[108].

Multi-layered evidence can be unified within a conceptual framework linking genotype, molecular phenotype, cellular and zonation phenotypes, clinical phenotype, and therapeutic targeting. PNPLA3 I148M illustrates this hierarchy: the risk allele (genotype) impairs triglyceride hydrolase activity and causes mutant protein accumulation on lipid droplets (molecular phenotype)^[23], preferentially affecting periportal hepatocytes as demonstrated by single-cell eQTL mapping (cellular and zonation phenotype)^[54], which manifests as accelerated MASH and fibrosis progression (clinical phenotype)^[24], and is now being addressed by allele-specific siRNA and ASO therapeutics (therapeutic targeting)^[93]. The same framework applies in reverse to protective alleles, such as HSD17B13 loss-of-function, where the therapeutic goal is the pharmacological phenocopy of a beneficial genotype^[89]. In real-world clinical decision-making, these layers interact dynamically. Defining baseline susceptibility; epigenetic and environmental modifiers shape current disease activity; and the integration of multi-omics data with non-invasive fibrosis assessment guides treatment selection. For instance, a biopsy-proven MASH F2 patient who is a PNPLA3 G/G homozygote with concomitant obesity may be prioritized for combination therapy with metabolic agents (e.g., resmetirom, GLP-1 receptor agonists) and considered for enrollment in genotype-targeted siRNA/ASO trials.

Given MASLD’s complex, multifaceted pathophysiology, the transition from conventional “one-size-fits-all” treatment to precision medicine integrating patient-specific environmental, genetic, and epigenetic data is imperative. The convergence of genomic, transcriptomic, spatial multi-omics, and AI-based approaches now provides a multi-layered framework for this transition. Endotype-stratified approaches - using multi-omics integration with AI - have facilitated the identification of distinct molecular phenotypes (lipid-export failure, lipotoxic-inflammatory, fibrogenic, and cancer-prone trajectories) that conventional clinical phenotyping could not. Incorporating common risk alleles (PNPLA3, TM6SF2, GCKR, MBOAT7, HSD17B13, MARC1), DNA methylation modules, miRNA networks, and circulating protein/metabolite panels into unified risk engines might enable the prediction of rapid progression, extrahepatic complications, and pharmacologic responsiveness.

Biomarker development must pivot from static, largely anatomical readouts toward dynamic, mechanistically anchored indicators directly linked to therapeutic targets. Genetic and epigenetic markers - such as ancestry-calibrated PRS, PNPLA3/TM6SF2/MBOAT7/HSD17B13/MARC1 combinations, complement-related methylation signatures, ATAC-seq-derived chromatin accessibility biomarkers (e.g., ETS family TF binding sites, ZNF384, IKZF1), DNMT1-dependent methylation modules, and miRNA panels - are attractive because they inform baseline risk, pathway dominance, and pharmacologic leverage points simultaneously^[62,64].

The advancement of MASLD precision medicine depends on integrating multi-omics data with advanced computational tools to resolve biologically meaningful endotypes invisible to routine clinical phenotyping. High-dimensional genomics, epigenomics, transcriptomics, proteomics, and metabolomics, when combined using AI and machine learning, can derive latent “molecular signatures” that distinguish disease trajectories rather than treating MASLD as a monolithic entity. For example, the integration of ATAC-seq chromatin accessibility data with DNA methylome profiling and RNA-seq transcriptomics - as demonstrated in Korean MASLD cohorts - enables identification of stage-specific regulatory elements and TF networks (e.g., HNF4α, PPARα, ETS family) that are epigenetically silenced during disease progression^[62,64]. Similarly, combining genetic variant information (PNPLA3, TM6SF2, MARC1) with proteomic pathway analysis (EMT, hypoxia, EGFR signaling) provides mechanistic granularity for patient stratification beyond conventional histological scoring^[97,98]. Such systems-biology frameworks incorporating common risk alleles, DNA methylation modules, miRNA networks, and circulating biomarkers into unified risk engines will enable prediction of rapid progression, extrahepatic complications, and pharmacologic responsiveness with unprecedented precision.

A critical frontier is the extension of eQTL mapping from bulk and single-cell levels to spatially resolved analyses. Our ongoing spatial-eQTL projects employing spatial transcriptomics on liver biopsies aim to systematically map genotype-expression associations across 13 distinct lobular zones, thereby identifying therapeutic targets whose regulatory activity is confined to specific microanatomical compartments. This spatial regulatory atlas, combined with organoid-derived liver-on-a-chip validation systems, is expected to provide mechanistic foundations for spatially targeted precision therapeutics^[115]. Beyond research-stage multi-omics integration, AI is increasingly deployed in concrete clinical applications for MASLD. A landmark example is the AI-based digital pathology platform AIM-MASH, which has demonstrated superior performance over unassisted pathologist reads for inflammation, ballooning, and MASH resolution scoring in regulatory-grade validation studies and was recently qualified by the U.S. FDA as the first AI-based drug development tool to support histologic endpoint assessment in MASH clinical trials^[116]. AI-augmented magnetic resonance imaging-derived proton density fat fraction (MRI-PDFF) and elastography are also expanding the toolkit for non-invasive longitudinal monitoring of hepatic steatosis and fibrosis, complementing biopsy-based assessment^[117]. Multimodal AI models that jointly integrate genomic, transcriptomic, histological, and imaging data have emerged as the foundation for automated endotype classification and individualized treatment selection in routine hepatology practice^[114].

Clinical trial design for MASLD must shift from “all-comers” recruitment to genetically and endotype-enriched cohorts. Up-front stratification with PRS, key single-gene variants, and body-composition indices (including sarcopenia assessment) enables more efficient trial conduct. Adaptive platform trials incorporating real-time biomarker feedback could dynamically reassign non-responders to alternative mechanism-based arms, reduce the required sample sizes for meaningful effect detection, and accelerate combination regimen decision-making. Embedding multi-omics profiling and dynamic biomarkers into trial eligibility, stratification frameworks, and endpoints will be critical for translating proof-of-concept nucleic-acid and epigenetic therapies into scalable, personalized treatment algorithms implementable in routine MASLD care.

Beyond therapeutic development, precision medicine enables earlier disease interception through risk stratification of asymptomatic individuals. Population-based genetic screening using PRS can identify high-risk individuals for targeted imaging surveillance and preventive interventions (lifestyle modification, pharmaceutical prevention) before the development of advanced fibrosis, thus fundamentally shifting MASLD management from reactive to proactive paradigms.

Whole-genome sequencing, snRNA-seq, and spatial transcriptomics remain confined to research-intensive centers because of their high cost and infrastructure requirements. For routine clinical use, targeted genotyping panels covering the major risk and protective variants (PNPLA3, TM6SF2, GCKR, MBOAT7, HSD17B13, MARC1) offer a pragmatic and substantially less costly alternative, capturing the majority of clinically actionable common variants while minimizing data management and incidental-finding burdens^[109].

Most current MASLD PRS are derived from cohorts of European ancestry and demonstrate reduced predictive performance in non-European populations, with the greatest loss of accuracy observed in individuals of African ancestry^[118]. Methodological advances such as multi-ancestry meta-analyses and Bayesian frameworks leveraging shared cross-ancestry effects have improved portability^[119]. Ancestry-diverse GWAS reference datasets and standardized cross-population PRS validation will be necessary for equitable clinical implementation.

Clinical implementation of genetic testing for MASLD requires distinct considerations^[120]. Although the major MASLD risk variants are common polymorphisms with modest individual effect sizes, appropriate informed consent and genetic counseling infrastructure remain essential, particularly when testing is a part of broader genomic panel that may yield incidental findings. Protection against genetic discrimination varies substantially across jurisdictions and many countries, including those in East Asia, lack comprehensive legislation. Moreover, the regulatory pathways of the PRS-based clinical decision-making tools remain unclear.

Effective use of genetic and biomarker data in routine practice requires (i) electronic health record-embedded clinical decision support that translates genotype, PRS, and non-invasive fibrosis scores into actionable risk stratification at the point of care; (ii) structured education to enhance the genomic literacy of hepatologists and primary care physicians; and (iii) multidisciplinary care models combining hepatology, endocrinology, and clinical genetics expertise. The 2024 EASL-EASD-EASO guidelines provide an initial clinical entry point by endorsing referral for genetic evaluation in patients with a strong family history of severe liver disease or early severe phenotypes^[109].