AstraZeneca uses BostonGene AI foundation model to predict patient safety and efficacy in early oncology trials

Early oncology clinical trials suffer high failure rates due to difficulty predicting which patients will respond to treatment and identifying safety signals before they emerge. AstraZeneca needed tools to de-risk drug development and make more informed go/no-go decisions in early-phase trials.

AstraZeneca partnered with BostonGene to apply its AI foundation model for tumor and immune biology. The platform combines pre-trained foundation models with multi-modal data analytics—including cell-free RNA and tumor microenvironment profiling—to predict response dynamics, tolerability, and safety/efficacy outcomes, supporting biomarker development and trial design optimization.

The collaboration aims to accelerate clinical development timelines and reduce risk through predictive intelligence on patient outcomes. No quantitative results have been reported as the partnership was announced in January 2026 and is in early stages.

• Foundation AI models trained on tumor biology can predict treatment response and safety signals earlier in the drug development process.
• Multi-modal data integration (cell-free RNA, tumor microenvironment profiling) enables more precise patient stratification in oncology trials.
• AI-driven biomarker development allows sponsors to design safer and more effective therapies from the start of development.