Trans-omics gene-smoking interactions: a new framework for stratifying smoking-related lung cancer risk

Lung cancer remains the leading cause of cancer mortality worldwide, accounting for over 1.8 million lives annually.1 While tobacco smoking represents the most important and prevalent risk factor, there is a long-standing paradox in that ∼75% of lung cancers have a history of smoked but only 10%-15% of lifelong smokers are diagnosed with lung cancer.2 This paradox in lung cancer susceptibility has been a major driver of molecular epidemiology research over the last several decades to identify inherited genetic factors that would reveal “higher risk” smokers more likely to be diagnosed with lung cancer. In this issue of the American Journal of Respiratory and Critical Care Medicine, Xie et al3 leveraged consortium-scale data to conduct a novel, and potentially transformative, trans-omics gene-smoking interactions analysis elucidate this paradox to facilitate risk stratification and identify biomarkers that modify effects of smoking on lung cancer risk. By integrating genome-wide association study (GWAS) data from over 27 000 lung cancer cases and 450 000 controls with multiple molecular layers spanning the epigenome, transcriptome, proteome, and metabolome, the authors provide a new framework for understanding and stratifying smoking-related lung cancer risk (Figure 1). Conventional molecular epidemiology studies have revealed numerous lung cancer susceptibility loci, most notably the chromosome 15q25 region containing nicotinic acetylcholine receptor genes.4 The cumulative body of GWAS-related research has successfully identified genetic factors significantly associated with lung cancer susceptibility, with some loci refined to specific subgroups (eg, sex, ethnicity, smoking status, histologic subtypes, etc.); however, there has been very limited research on integrating numerous molecular modalities (viz layers) and conducting ­detailed gene-smoking interactions. Xie et al. break new ground by introducing “xWAS” (cross-omics wide association study), which is a comprehensive framework to test gene-environment interactions using multiple molecular layers simultaneously. This innovative approach leverages existing quantitative trait loci (QTL) databases to impute molecular trait values based on genotype, circumventing the prohibitive costs and efforts of directly measuring DNA methylation, gene expression, or protein levels in hundreds of thousands of participants. Using LDpred2-based polygenic scores, genetically predicted values were constructed for over 100 000 molecular biomarkers, focusing on traits with sufficient heritability to ensure robust downstream analyses.5 This strategy enabled interrogation of approximately 150 000 molecular traits for interaction with smoking exposure, representing the largest search space ever examined for gene-smoking interactions. Despite casting a “wide-net” in terms of the number of data inputs to analyze, the rigorous 2-phase discovery and replication analysis distilled the findings to 8 molecular markers showing statistically robust gene-smoking interactions (discovery P ≤ .05 with successful replication). Remarkably, 7 of markers were DNA methylation sites (CpG sites), with 4 residing in the 15q25.1 region harboring the CHRNA3/5-CHRNB4-IREB2 nicotinic receptor cluster, 2 mapping to RDM1 on 17q12, and 1 to GTPBP1 on 22q13.1. The sole expression trait was RP11-326C3.14 at chromosome 11p15.5. While these markers do not appear to contribute as main effects on lung cancer risk, rather they modify risk in the presence of smoking. The interaction patterns were striking, with some variants showing minimal impact among never smokers but amplified risk among heavy smokers. This finding suggests they are effect modifiers that reveal statistical and biological underpinnings in the presence of tobacco exposure.6 The study also includes a potentially clinically actionable Molecular Modifying Score (MMS), which is a metric that quantifies how strongly an individual’s molecular profile amplifies the carcinogenic effects of smoking. Conceptually analogous to a polygenic risk score, but the MMS explicitly encodes interaction effects and aggregates weighted contributions from 5 relatively independent biomarkers. Lung cancer risk stratification was observed among ever smokers by MMS tertiles. Heavy smokers in the highest tertile exhibited markedly higher lung cancer risk than heavy smokers in the lowest tertile. The predicted lung cancer probabilities diverged substantially between MMS groups, demonstrating that 2 individuals with identical smoking histories could have vastly different risk trajectories based on their molecular architecture.7 While the MMS yielded only modest improvement in overall discrimination (integrated area under the curve increased from 0.839 to 0.842), this incremental gain is meaningful within the smoking population. The MMS is not a generalized lung cancer risk score but rather captures additional risk attributable to molecular contributors, which could be critical for precision prevention and risk assessment. Beyond risk stratification, the trans-omics integrative analysis revealed interesting biological insights. The authors identified 55 CpGs correlated with 26 gene expressions in cis, and notably, NELFE (negative elongation factor E), showing trans-acting correlations with 3 downstream proteins, including ANAPC7 (anaphase-promoting complex subunit 7), which links to ­metabolite profiles. NELFE lies in the MHC class III region, implicating immune and inflammatory pathways as potential mediators of smoking-related carcinogenesis. The authors mapped multiple putative cross-omics relationships that form molecular pathways, which may mediate the impact of smoking, converging on processes including cell proliferation and DNA repair. While these network inferences are correlational, they generate concrete mechanistic hypotheses amenable to experimental validation, aligning with broader functional genomics efforts in lung cancer.8 The many strengths of this study include exceptional scale through international collaborations, rigorous replication design, and innovative use of genetically imputed molecular traits that reduce reverse-causation concerns. By explicitly modeling gene-smoking interactions rather than treating smoking as a simple covariate, the work addresses a key source of “missing heritability” in lung cancer susceptibility.9 However, there are limitations that warrant consideration. The analyses capture only the heritable component of molecular phenotypes, missing non-genetic influences, including direct epigenetic effects of smoking, even though smoking is known to induce widespread methylation changes associated with future lung cancer risk.8 The restriction to European ancestry populations limits generalizability, particularly given evidence of population-specific genetic architecture in lung cancer risk.10 Additionally, blood-based QTL resources may miss lung tissue-specific regulatory effects critical to carcinogenesis. Current lung cancer screening eligibility relies primarily on age and smoking history, while lung cancer can vary markedly among those eligble.11 Incorporating molecular modifiers, such as MMS, into screening guidelines could refine the selection of high-risk smokers optimizing the balance of benefit and harm. Knowledge of heightened genetic susceptibility might also enhance motivation for smoking cessation among vulnerable individuals.7 The identified molecular pathways have potential therapeutic targets. Genes showing strong smoking interactions may encode druggable proteins targetable for chemoprevention. Understanding how smoking modifies treatment response based on tumor genetics could inform precision oncology approaches.12-14 This trans-omics study presented by Xie et al. could present a watershed moment and a new approach in understanding gene-environment interactions in not just lung cancer, but other smoking-related cancers. By demonstrating that multiple molecular layers fundamentally modify the carcinogenic influence of smoking on cancer risk, this work provides both biological insights and practical tools for real-world risk stratification. The MMS offers a tangible metric for identifying vulnerable smokers, while the cross-omics analyses generate testable hypotheses about carcinogenic mechanisms. As trans-omics resources expand and diversify, extending this framework to non-European populations, directly measured lung tissue omics, and experimental validation will be critical. This study moves us closer to a future where lung cancer prevention and early detection can be tailored not only to smoking exposure but also to individual molecular susceptibility, whereby transforming population-based screening into precision prevention.15 The framework for trans-omics gene-smoking interactions. This graphical schematic summarizes the strategy to identify trans-omics gene-smoking interactions to risk-stratify smokers by integrating consortium-scale GWAS data with xQTL resources for DNA methylation, gene expression, proteins, and metabolites alongside detailed smoking exposure (pack-years, intensity, and duration). The xWAS pipeline uses imputed molecular data to test gene-smoking interactions for lung cancer risk, which revealed 8 biomarkers (7 CpG sites and 1 gene expression trait). These biomarkers were combined into a Molecular Modifying Score (MMS), which is a weighted sum of 5 relatively independent markers to quantify smoking-related risk and stratifies ever smokers into low, medium, and high susceptibility groups. The framework illustrates how MMS-informed risk stratification can potentially refine the selection of high-risk smokers for lung cancer screening and generate hypotheses about cross-omics pathways for precision prevention and early detection. Supplementary material is available at American Journal of Respiratory and Critical Care Medicine online. Please see the ICMJE disclosure forms, which have been provided as supplementary material.