AI-driven multi-omics integration in precision oncology: bridging the data deluge to clinical decisions
Chou‐Yi Hsu, Shavan K. Askar, Samer Saleem Alshkarchy, Priya Priyadarshini Nayak, Kassem Al Attabi, Mohammad Ahmar Khan et al. · 2025 · 128 citationsRead the paper
Cancer's staggering molecular heterogeneity demands innovative approaches beyond traditional single-omics methods. The integration of multi-omics data, spanning genomics, transcriptomics, proteomics, metabolomics and radiomics, can improve diagnostic and prognostic accuracy when accompanied by rigorous preprocessing and external validation; for example, recent integrated classifiers report AUCs around 0.81-0.87 for difficult early-detection tasks. This review synthesizes how artificial intelligence (AI), particularly deep learning and machine learning, bridges this gap by enabling scalable, non-linear integration of disparate omics layers into clinically actionable insights. We explore cutting-edge AI methodologies, including graph neural networks for biological network modeling, transformers for cross-modal fusion, and explainable AI (XAI) for transparent clinical decision support. Critical applications are highlighted, such as AI-driven therapy selection (e.g., predicting targeted therapy resistance), proteogenomic early detection, and radiogenomic non-invasive diagnostics. We further address translational challenges: data harmonization, batch correction, missing data imputation, and computational scalability. Emerging trends, federated learning for privacy-preserving collaboration, spatial/single-cell omics for microenvironment decoding, quantum computing, and patient-centric "N-of-1" models, signal a paradigm shift toward dynamic, personalized cancer management. Despite persistent hurdles in model generalizability, ethical equity, and regulatory alignment, AI-powered multi-omics integration promises to transform precision oncology from reactive population-based approaches to proactive, individualized care.
What the full text shows
Working prototype- Integrated omics classifiers for early-detection tasks — AUC 0.81-0.87 (Difficult early-detection tasks)
- Graph-based attention models (moGAT) for cancer subtype classification — AUC 0.92 (Pan-cancer cohort studies)
- Gradient Boosting models for leukemia subtyping — 97% accuracy using 17 features (Limited samples with clinical and cytogenetic features)
- Variational Autoencoders (VAEs) for single-cell multi-omics fusion — 15% higher accuracy vs. linear methods (Integrating DNA methylation with RNA-seq data)
- Similarity Network Fusion (SNF) for survival prediction in neuroblastoma — 12% improvement over simple concatenation (Fusing patient similarity networks across omics layers)
- Transformer-based architectures for inferring tumor clonality — 7-10% improvement in accuracy (Genome-wide integration and imputation of missing proteomic data)
- Setup
- Synthesis of existing research including datasets such as TCGA, ROSMAP, LGG, and BRCA.
- How
- The paper is a narrative review synthesizing AI methodologies (GNNs, Transformers, VAEs, and classical ML) for integrating genomics, transcriptomics, proteomics, metabolomics, and radiomics. It evaluates data-fusion strategies (early, late, and hybrid) and explainability tools like SHAP and Integrated Gradients.
- Limitations
- Narrative review susceptible to selection bias (not a systematic PRISMA-style review) · Lack of large-scale prospective randomized studies for clinical outcomes · Model generalizability across diverse healthcare systems and populations · Computational scalability for billion-parameter models
Summarized by Gemma 4 from the open-access full text · 11 Oct 2026. Read it