Brian Pierce, UMD CBMG & IBBR
High resolution modeling of immune recognition with deep learning
Abstract:
Antibodies and T cell receptors (TCRs) are critical components of adaptive immunity, with many implications for disease research and increasing use as therapeutics. Advances in deep learning, including the development of AlphaFold, provide an opportunity to address major challenges, including accurate modeling and design of antibody and TCR recognition. Recently we developed the algorithm and web server TCRmodel2, which is able to model TCR complexes with peptide-MHC targets from sequence, through several optimizations of AlphaFold2. To provide insights into the performance of AlphaFold for modeling immune recognition, we benchmarked AlphaFold2 with massive sampling, which was reported to improve accuracy with some complexes, in comparison with default AlphaFold2 and AlphaFold3, for antibody-protein, antibody-peptide, and TCR-peptide-MHC complexes. We found that massive sampling improved modeling accuracy in AlphaFold2, with increases in benchmark success rate for all three interaction classes versus the default protocol, while AlphaFold3 generally outperformed default AlphaFold2. Scores from AlphaFold were found to be strongly associated with correct modeling performance, providing useful indicators for predictive modeling, while overall limitations in success indicate that additional modeling developments are still needed.
Speaker: Brian Pierce, University of Maryland Department of Cell Biology and Molecular Genetics & the Institute for Bioscience and Biotechnology Research
Dr. Brian Pierce is an Associate Professor at the University of Maryland in the Department of Cell Biology and Molecular Genetics and the Institute for Bioscience and Biotechnology Research. After receiving his Ph.D. in Bioinformatics from Boston University, he worked at Pfizer as a Senior Scientist and the University of Massachusetts Medical School as a research faculty member before joining the University of Maryland in 2014. His laboratory’s research area is computational structural biology, with a particular focus on immunology, including modeling and design of T cell receptors, antibodies, and vaccines. The Pierce lab has generated numerous tools and algorithms, including TCR3D, a database of T cell receptor structures, CoV3D, a database of coronavirus and antibody structures, and TCRmodel, an algorithm and web server to model T cell receptors and their complexes from sequence.
Host: Jeffery Klauda
Seminars start at 4:00 pm, and refreshments will be served at 3:45 pm. All seminars are held in the Conference Room (1116) of the Institute for Physical Science and Technology (IPST) Building (Bldg #085) unless otherwise noted.