MSU AI model finds drug candidates for liver cancer, lung disease
Get the Health newsletter
Daily health & science — research, biotech, public health, the studies worth knowing. Free.
- Michigan State University-led team developed the GPS deep-learning platform, trained on millions of published chemical-gene expression measurements, to predict a compound's effect on specific genes based solely on its chemical structure, per a study published in Cell.
- For hepatocellular carcinoma (HCC), the third leading cause of cancer-related death worldwide, the platform identified two new compounds that reduced tumor size when tested in mice, with collaborators from Stanford's Asian Liver Center.
- For idiopathic pulmonary fibrosis (IPF) — a chronic lung disease with a three-year median survival rate and no curative options — the team identified one repurposed drug and two new compounds, validated in mice and in live cultures of human lung tissue from Corewell Health's transplant program.
- Xiaopeng Li, an MSU associate professor and study co-author, said the AI component helped probe IPF 'differently and more systemically' after 20 years of failed efforts to find new drugs for the disease.
- The project brought together over 20 researchers across computational biology, chemistry, and clinical medicine, with co-senior author Jiayu Zhou (now at the University of Michigan) collaborating with MSU's Bin Chen on the model.
- The team has released its code and a public web portal for virtual compound screening, enabling other researchers to apply GPS to additional diseases beyond HCC and IPF.
Why it matters: By releasing the GPS platform's code and web portal publicly, MSU has turned a single drug-discovery result into a reusable screening tool that other labs can apply to different diseases, potentially compressing the years-long compound-screening process. The IPF application is especially notable given two decades of failed drug programs for the disease, and the study's end-to-end pipeline — from AI prediction to mouse validation to human tissue testing — offers a replicable template for AI-driven therapeutic development.
Ask SkimNews
