Single-Cell Mendelian Randomization Identifies Cell-Type Specific Genetic Drivers of Lung Cancer Subtypes

BackgroundGenome-wide association studies (GWAS) have identified many loci linked to lung cancer, but connecting these loci to causal genes and relevant cell types remains difficult. Traditional Mendelian randomization (MR) using bulk tissue eQTLs averages signals across diverse cells, masking cell-specific effects. Here, we apply single-cell eQTL based MR to infer causal relationships at immune-cell resolution, revealing mechanisms not detectable in bulk analyses. MethodsWe conducted single-cell cis-eQTL MR across multiple immune cell subtypes to identify genes causally linked to lung cancer subtypes, including non-small cell lung cancer, further classified into adenocarcinoma and squamous cell carcinoma, and small cell lung cancer. Sensitivity and colocalization analyses confirmed shared genetic variants driving both expression and cancer risk. Bulk-tissue MR was conducted to see if the single-cell positive genes also show significance in whole blood and lung tissue. Phenome-wide MR (PheMR) further assessed pleiotropic associations across complex traits. ResultsSingle-cell Mendelian Randomization (scMR) uncovered 375 cell-type specific causal links between gene expression and lung cancer subtypes. Colocalization confirmed shared causal variants in 35 gene-cell combinations. Cross-tissue validation highlighted five consistent genes--HSPA1B, HLA-DOA, GPX1, ZP3, and CORO1B. PheMR revealed these genes additional associations with other diseases, suggesting broader functional relevance. ConclusionsThe scMR workflow provides a high-resolution causal map of lung cancer susceptibility within immune contexts. The results nominate HSPA1B, HLA-DOA, GPX1, ZP3, and CORO1B as promising candidates for mechanistic and translational follow-up, demonstrating the power of single-cell causal inference in uncovering disease mechanisms obscured in bulk analyses.