BIOE 500 Spring 2026 - Dr. Michael Robben

Leveraging AlphaFold to predict TCR-pMHC specificity in autoimmune disease

Abstract: Inflammatory autoimmune diseases like Type 1 Diabetes are driven by poor tolerization of the T cell receptor to pancreatic antigens. Determining why T cells malfunction in this disease is complicated by the enigmatic nature of TCR-pMHC interactions. By being able to predict which T cells target self-antigens, we can improve our understanding of why these cells malfunction in disease. We explored structural prediction algorithms for their ability to generate realistic quaternary TCR-pMHC conformations, and the ability for structural features to predict interactions. We determined that AlphaFold produced the highest quality multimeric structures while generalizing to unseen epitopes. Structure quality was not a predictor of interaction but did produce features characteristic of interaction. Machine learning and deep learning models predicted interaction at higher accuracy than sequence based models and highlighted features important for discerning interacting from non-interacting structures. Molecular dynamics simulation revealed that non-interacting structures stabilize slower than interacting structures.

Biography: Dr. Robben is a computational immunologist working in the department of Animal Science at the University of Illinois. His work focuses on understanding how the environment changes the risk for individuals to develop autoimmune disease and cancer. Dr. Robben graduated with a B.S in Biology from Salisbury University in 2016 and a Ph.D. in Molecular Biology from South Dakota State University in 2022. Dr. Robben is dedicated to advocating for the use of machine learning and bioinformatics to biological research and teaches a class focused on biological image analysis.