AI advancements improve bladder cancer detection and drug discovery

Researchers have developed two distinct artificial intelligence models aimed at advancing medical diagnostics and pharmaceutical development.

At the University of Plymouth, a team led by Professor Shang Ming Chu developed PRECISE-AGZ, an AI model capable of identifying patterns associated with bladder cancer. By analyzing electronic health records from approximately 70,000 patients between 1995 and 2020, the model identified 38 key health indicators. In testing, the model detected 85% of cancer cases and correctly identified 91% of non-cancerous cases. The system showed the ability to capture signals related to the disease up to five years before official diagnosis, though its highest efficacy was noted within the 12 months preceding diagnosis.

Separately, researchers at New York University have introduced Tautomer-Predictor, a deep learning model designed to accelerate drug discovery. The tool predicts the most stable chemical shapes of drug-like molecules by identifying the precise position of hydrogen atoms. This addresses the complex chemical issue of tautomers, where hydrogen atoms shift positions within a molecule. The model can analyze approximately 4.6 million chemical compounds in just 3.2 hours using a single computing node. The research team has released the model as open-source to assist research institutions in screening compounds before costly development stages.