Predictive Modeling & AI

Developing CAR-T cell therapy, a highly personalized cancer treatment, is notoriously slow and expensive, taking 8 to 12 years with a failure rate of up to 60%. Researchers have created the Bio AI Agent, a system of collaborative artificial intelligence programs powered by large language models. These digital agents work together to automatically discover targets on cancer cells, predict potential toxic side effects, and design optimal molecules. By automating these complex, manual stages of drug design, the system aims to bypass traditional bottlenecks, lowering the high attrition rates and bringing life-saving cancer treatments to patients much faster.

To control infectious diseases like COVID-19, public health officials rely on mathematical numbers called R0 and Rt, which measure how fast a virus is spreading. However, early calculations often struggle to account for the exact timing of when people become contagious, especially when they spread the virus before showing symptoms. Researchers developed a refined mathematical model that uses a specific bell-curve distribution to map these transmission intervals. By better accounting for presymptomatic spread, this framework provides a more accurate tool for predicting outbreaks, evaluating the success of lockdowns, and determining exact vaccination rates needed to protect communities.

Modern hospitals generate massive amounts of patient data, from medical images and genetic sequences to written health records, but these formats rarely talk to each other. Monash University is building Australia's first healthcare-specific AI foundation model to solve this fragmentation. Instead of using separate tools for different tasks, this single AI is designed to analyze diverse, multimodal patient data at a massive scale. By connecting the dots between a patient's scans, DNA, and medical history, the system aims to help doctors make faster, highly accurate diagnoses and design personalized treatment plans.