Machine Learning and AI-Driven Multi-Omics Biomarkers of Cancer Therapy Response and Resistance

Despite an expanding arsenal of cancer therapies, clinicians still cannot reliably predict which patients will benefit from a given treatment. Response is governed by molecular alterations distributed across the genome, epigenome, transcriptome, proteome, and metabolome, and single-layer assays capture only a fraction of this complexity. Multi-omics profiling now offers a systems-level portrait of the tumor and its microenvironment, and artificial intelligence has become the engine for integrating these high-dimensional layers into predictive signatures. Yet two gaps stall progress: many AI models are opaque, cohort-specific, and difficult to validate, while most candidate biomarkers are never functionally tested or carried to clinical endpoints. The result is an abundance of published signatures and a scarcity of usable ones. This Research Topic focuses on biomarkers that are AI-derived yet interpretable, rigorously validated, and actionable capable of predicting therapy response and resistance in real-world care.The problem we aim to tackle is the persistent gap between biomarker discovery and clinical impact, a gap widened, not closed, when high-performing but opaque AI models cannot be validated or trusted. Enormous numbers of multi-omics signatures have been published, yet few influence treatment decisions because they are underpowered, poorly validated, disconnected from biology, or impossible to interpret. This Research Topic seeks studies that move biomarkers along the full pipeline from rigorous, interpretable discovery in well characterized cohorts, through mechanistic and external validation, to demonstration of clinical actionability. Recent advances make this achievable: explainable AI methods can now expose the molecular logic behind predictions; transcriptomics, single-cell and spatial omics resolve the cellular states that govern response; longitudinal profiling tracks resistance as it emerges; and harmonized public cohorts enable genuine cross population testing. By prioritizing transparency, validation, and translation over discovery alone, we hope to surface biomarkers that can stratify patients, anticipate resistance, and guide therapy selection, advancing precision oncology from promise to practice.We welcome Original Research, Method and Technology reports, Reviews, Mini-Reviews, and Brief Research Reports that develop, validate, or translate AI-driven multi-omics biomarkers of treatment response and resistance, spanning preclinical models, clinical cohorts, and computational frameworks. Specific themes include:• Integrative multi-omics signatures predicting response to targeted therapy, immunotherapy, chemotherapy, or radiation• Explainable and interpretable AI linking multi-omics features to biological mechanism• Models that generalize across cohorts, platforms, and institutions, with rigorous external validation• Longitudinal and on-treatment profiling that captures adaptive resistance and minimal residual disease• Functional and mechanistic validation connecting candidate biomarkers to therapeutic vulnerability• Translation toward clinically actionable, deployable assays and decision-support tools• Computational and structural analysis using bioinformatics, transcriptomics, statistical modeling, and AI-driven methods to connect high-throughput molecular profiles with therapy response and mechanism Topic editors: Anirudh Gaur, Dr. Vikas Soni Source: https://www.frontiersin.org/research-topics/83635/machine-learning-and-ai-driven-multi-omics-biomarkers-of-cancer-therapy-response-and-resistance