IPM Take
This is not a drug result, and it should not be written like one. The point is the evidence machinery behind future treatment selection. AI multiomics can help identify biomarkers, predict response, map resistance and stratify patients for trials. The risk is complexity without action. The opportunity is smarter evidence generation, if the outputs are clinically interpretable and validated.
Executive Summary
BostonGene announced nine abstracts accepted for ASCO 2026, focused on AI-powered models and multimodal profiling in tumour and immune biology. The accepted studies included use of the Tumor Portrait test and Kassandra cell deconvolution, spanning tissue and peripheral blood analysis. The company described applications in biomarker discovery, immunotherapy response prediction, toxicity risk, resistance mapping and clinical trial stratification across solid tumours and hematologic malignancies.
Why it matters
- Researchers / academia: AI multiomics could improve understanding of response, resistance and tumour microenvironment biology.
- Industry / innovation partners: Better stratification may de-risk trials and identify patients most likely to benefit.
- Clinicians / patients: The long-term value depends on whether complex molecular outputs become usable treatment or trial decisions.
Precision oncology has created a data problem. DNA mutations matter, but they rarely tell the whole story. RNA expression, immune context, tumour microenvironment, blood-based immune signals and clinical outcomes all add layers of meaning.
AI multiomics is trying to connect those layers. At ASCO 2026, the focus was not one drug or one cancer. It was the use of tumour and immune biology models to discover biomarkers, predict response, identify toxicity risk and improve clinical trial stratification.
This is promising because many trials fail not only because the therapy is weak, but because the wrong patients are enrolled, resistance is poorly understood or biomarkers are too crude. A richer biological model could make development more precise.
The caution is equally important. Multiomics can become impressive but unusable if outputs are too complex, thresholds are unclear or clinicians cannot act on the report. For IPM, the question is whether AI multiomics can move from discovery layer to decision layer. That is where access begins.
