Story perspectives

UCSD Model Predicts Immunotherapy Response, Beats Competing Methods

5/27/2026

Summary

Story summary

UC San Diego researchers led by Professor Trey Ideker unveiled MutationProjector, a model predicting cancer treatment response.
It was trained on over 30,000 genomes from ten solid tumor types.
Fine-tuned on 94 samples, it outperformed other methods in forecasting immunotherapy response and identified resistance biomarkers KMT2A, SMARCA4, KRAS-STK11 and STK11-KEAP1.
Researchers plan to expand the dataset to about 300,000 genomes and add more cancer types and multimodal inputs.

1 of 1

Story summary

UC San Diego researchers led by Professor Trey Ideker unveiled MutationProjector, a model predicting cancer treatment response.
It was trained on over 30,000 genomes from ten solid tumor types.
Fine-tuned on 94 samples, it outperformed other methods in forecasting immunotherapy response and identified resistance biomarkers KMT2A, SMARCA4, KRAS-STK11 and STK11-KEAP1.
Researchers plan to expand the dataset to about 300,000 genomes and add more cancer types and multimodal inputs.