AI-guided multi-omics analysis identifies NPC1-modulated susceptibility to SARS-CoV-2 infection under PM2.5 exposure

Exposure to airborne fine particulate matter (PM_2.5) has been linked to increased risk of the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection, yet the underlying mechanisms remain unclear. Here, by leveraging a fine-tuned foundation model of single-cell transcriptomics, we uncover shared transcriptional signatures between PM_2.5 exposure and SARS-CoV-2 infection. We further validate this association using population-level epidemiological analyses and perform genome-wide association studies (GWAS) to identify genetic variants that modulate infection risk under PM_2.5 exposure. In addition, we identify NPC1 as a key modulator involved in SARS-CoV-2 infection efficiency under virus-laden PM_2.5 exposure through integrative functional genomic analyses and in vitro experiments. Our findings suggest that PM_2.5 facilitates viral entry through an NPC1-modulated endo-lysosomal pathway, providing a mechanistic explanation for observed pollution-related susceptibility. By integrating artificial intelligence (AI)-guided transcriptomics, epidemiology, GWAS, functional genomics, and in vitro verification, our study elucidates how environmental and genetic factors jointly influence SARS-CoV-2 susceptibility. This work highlights how AI-assisted multi-omics integration systematically decodes the health impacts of environmental exposures from molecular to population levels and informs air quality policy and infectious disease preparedness.