Oncogenomics-Driven Multimodal AI for Predicting and Explaining Immune Checkpoint Blockade Response

Immune Checkpoint Blockade (ICB) has reshaped cancer care and can deliver durable remission in malignancies such as melanoma and non-small cell lung cancer. However, most patients do not benefit, and current biomarkers (e.g., PD-L1 expression) remain insufficient for precise stratification, partly because immune responses are dynamic and systemic. Unimodal biomarkers cannot capture the interaction among tumor immunogenicity, the tumor microenvironment, and host immune status.This Research Topic invites work on oncogenomics-driven multimodal learning to improve the prediction, interpretation, and clinical translation of ICB response and resistance. We welcome studies integrating tumor and host genomic features (somatic variants, mutational signatures, copy-number alterations, HLA genotype/loss, neoantigen features, clonality, immune-evasion alterations) with immunogenomic readouts (bulk/single-cell transcriptomics, spatial/pathomics, epigenomics, TCR repertoire), and—where relevant—radiomics and longitudinal EHR data to support real-world deployment.Join us in moving beyond static, unimodal correlates toward validated, generalizable, and clinically actionable models that (i) identify robust genomic/immunogenomic biomarkers, (ii) uncover interpretable molecular mechanisms of ICB response and resistance, and (iii) translate across institutions and populations. We especially welcome methodologies that solve practical implementation barriers: utilizing Incomplete Multimodal Learning to handle missing data, deploying Federated Learning to overcome privacy silos, and ensuring clinical trust through Explainable AI.This Research Topic welcomes:• Multi-omics and integrative oncogenomics studies of ICB response/resistance (DNA/RNA/epigenome/TCR/spatial), with strong validation in independent cohorts• Methods for multi-modal fusion where genomics/immunogenomics are central drivers• Studies mapping somatic variation, mutational signatures, HLA/neoantigen features, and immune-evasion alterations to treatment outcomes• Federated learning frameworks enabling cross-site modeling of genomic and clinical data while preserving privacy.• Explainable models that provide gene/variant/pathway-level interpretations and hypotheses• Clinical or translational validation comparing genomic/immunogenomic AI biomarkers to standard-of-care metrics, ideally with prospective or multi-center evaluationSubmissions based purely on re-analysis of public datasets should include independent validation and, where feasible, experimental or orthogonal verification. Topic editors: Junwei Liu, Xiaowen Hu, Zhongxu ZHU, Lulu Liu Source: https://www.frontiersin.org/research-topics/77482/oncogenomics-driven-multimodal-ai-for-predicting-and-explaining-immune-checkpoint-blockade-response