Immunotherapy drug target identification using machine learning and patient-derived tumour explant validation

2026-05-18 · Source: Nature Machine Intelligence · Field: Science & Research — Life Sciences & Biology, Health & Medical Research, Mathematics & Computational Sciences · Depth: Expert, extended

Summary

Researchers introduce Mining Immunotherapy Drug tArgetS (MIDAS), a multimodal graph neural network (GNN) system designed for immuno-oncology target discovery. This system integrates diverse data, including gene interactions, multi-omic patient profiles, immune cell biology, antigen processing, disease associations, and phenotypic consequences of genetic perturbations. MIDAS demonstrates robust performance, outperforming existing computational baselines like OpenTargets and ranking approved targets higher than those in clinical development. It also successfully identifies immunotherapy-response-associated genes in previously unseen patients. Interpretability analyses highlight the system's reliance on autoimmunity, regulatory networks, and immuno-oncology pathways. Functional validation in TRACERx melanoma patient-derived explants showed that perturbing oncostatin M (OSM)–oncostatin M receptor (OSMR) signaling, a target proposed by MIDAS, reduced dysfunctional CD8+ T cells and CCL4 levels, supporting its role in modulating the tumor microenvironment towards immunosuppressive phenotypes.

Key takeaway

For AI scientists and research scientists engaged in drug discovery, MIDAS offers a validated framework to accelerate immuno-oncology target identification. You should consider adopting multimodal GNN approaches to integrate diverse biological data, as this method has demonstrated superior performance over traditional techniques and can de-risk expensive drug development programs by prioritizing more promising targets like OSM-OSMR signaling.

Key insights

MIDAS, a multimodal GNN, effectively identifies novel immuno-oncology targets by integrating diverse biological data.

Principles

• Multimodal data integration enhances disease profiling.
• GNNs effectively leverage gene interaction networks.
• Human genetic evidence improves drug development success.

Method

MIDAS uses a GNN on a multimodal biological graph, integrating patient omics, immune cell atlases, HLA-peptidomics, GWAS, and CRISPR screens to predict immuno-oncology targets.

In practice

• Apply GNNs for complex biological data integration.
• Validate ML predictions using patient-derived explants.
• Prioritize targets with strong human genetic evidence.

Topics

• MIDAS
• Graph Neural Networks
• Immuno-oncology Target Discovery
• Multi-omics Data Integration
• Patient-Derived Explants

Code references

• bmbolstad/preprocesscore
• czbiohub-sf/scell_lung_adenocarcinoma
• augustgm/MIDAS
• augustgm/MIDAS

Best for: AI Scientist, Research Scientist

Related on AIssential

• Graph Therapeutics brings total funding to over $10M to advance precision immunology — AI-powered analysis of live patient cells, multi-omics data, and functional biology decodes immune dysfunction for precision immunology.
• 🔬 Training Transformers to solve 95% failure rate of Cancer Trials — Ron Alfa & Daniel Bear, Noetik — Improving cancer drug success requires advanced AI models for precise patient selection based on multimodal biological data.
• Predicting Immune Biomarkers with MultiModal Mixture-of-Expert Pathology Foundation Models Empowers Precision Oncology — MixTIME integrates multimodal pathology foundation models via a Mixture-of-Experts architecture to predict immune biomarkers for precision oncology.
• Gene dependency-informed inference of response to targeted cancer therapies — FORGE co-models drug response and gene dependency to predict cancer therapy benefit, improving mechanistic understanding and prediction.
• Bridging the phenotype-target gap for molecular generation via multi-objective reinforcement learning — ExMolRL combines phenotype-guided generation with target-aware reinforcement learning for de novo drug discovery.
• Radical Numerics launches with $50M to build general biological intelligence — Multimodal AI models integrating diverse biological data can advance human health through diagnostics and drug discovery while simultaneously bolstering biosecurity.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Nature Machine Intelligence.