From Variant to Biomarker in NSCLC Immunotherapy Resistance: Multiomics Evidence Chains and Accountable AI Integration

Immune checkpoint inhibitors have become integral to the management of non-small cell lung cancer (NSCLC), yet both primary and acquired resistance remain frequent and are only partially captured by routine biomarkers such as programmed death-ligand 1 (PD-L1) immunohistochemistry and tumor mutational burden (TMB). Resistance is increasingly viewed as a multiaxis functional phenotype shaped by antigenicity and neoantigen quality, antigen processing and presentation competence, interferon signaling and adaptive resistance programs, tumor-immune spatial organization, suppressive myeloid/stromal ecosystems, and metabolic constraints that limit effector function. Multiomics profiling provides a practical route to translate genomic event anchors into reproducible, mechanistically interpretable biomarker outputs by assembling coherent evidence chains across genomics, transcriptomics, epigenomics, proteomics, and metabolomics, complemented by spatial assays, digital pathology, and imaging-derived surrogates.