Multi-omics integration deciphers arsenic-induced multi-organ toxicity and the novel ferroptosis axis

Chronic arsenic exposure threatens over 200 million people worldwide and induces multi-organ injury, yet the panoramic molecular reprogramming across organs remains incompletely understood, and traditional single-omics approaches fail to capture cross-level and cross-organ regulatory associations. This review systematically integrates evidence from global epidemiology to single-cell spatial omics, tracing the evolution of multi-omics technologies-from single-platform profiling to data fusion strategies such as coupled matrix factorization (CMF) and the DIABLO framework, two complementary multi-omics integration approaches-and to cutting-edge spatial transcriptomics. We highlight ferroptosis as a common mechanism in arsenic-induced multi-organ injury. At the molecular level, we propose a mechanistic model wherein arsenic (AsIII) disrupts selenium metabolism by inhibiting Sec-tRNA^Sec synthesis, thereby impairing selenoprotein (especially GPX4) biosynthesis, a paradigm distinct from the traditional "ROS burst → lipid peroxidation" theory. Supporting evidence includes reduced ⁷⁵Se incorporation into cellular RNA, genetic deletion of PRDX6 exacerbating ferroptosis, and rescue by selenium supplementation via Nrf2 activation. At the organ level, we compare toxicity features and propose that tissue-intrinsic ferroptosis thresholds-determined by iron content, PUFA-phospholipid composition, GSH reserves, GPX4 redundancy, and selenium availability-govern differential organ susceptibility; the brain shows extreme vulnerability, whereas the liver exhibits relative resistance. Emerging spatial omics further reveals elevated arsenic-responsive gene expression in tumor-adjacent regions. This systems toxicology paradigm offers mechanistic grounding for combinatorial biomarker panels and genotype-guided precision selenium supplementation, although definitive causal validation through tissue-specific Gpx4 knockout or Sec-tRNA^Sec rescue experiments remains a critical next step. We also discuss prospects for AI-driven toxicity prediction models.