Seminar: Bridging the Resolution Gap in Computational Modeling of Protein 3D Structures
Assistant Professor, Auburn University
Wednesday, March 10, 2021
10:00am - 11:30am
Computational modeling remains the only truly scalable solution to derive atomic-level 3D protein structures at genomic scale when experimental structure determination techniques are not feasible or practical. Despite the dramatic recent progress in protein structure prediction, the qualities of computationally predicted protein models still vary considerably depending on the availability of evolutionary information, resulting in a resolution gap in computational modeling of protein 3D structures. Bridging this gap is crucial in order to make computational protein modeling practically useful in biomedical and life science research.
In this talk, I will present my group’s work in developing novel computational and data- driven methods powered by deep learning for high-resolution computational protein modeling, paving the way to near-experimental accuracy in protein structure prediction. First, I will introduce a supervised learning framework for estimating the quality of a predicted protein model using ensembles of Deep Residual Neural Network (ResNet) classifiers. Then, I will present a high-resolution structure refinement method that minimizes fine-grained structural anomalies estimated from Deep Conditional (Markov) Neural Fields (DeepCNF) via transformed restrained energy minimization in order to drive moderately accurate protein models towards high-resolution. Finally, I will outline future research directions on attaining atomic-level accuracy in computational protein modeling at genomic scale.
Debswapna Bhattacharya is an Assistant Professor in the Department of Computer Science and Software Engineering at Auburn University. He received his Ph.D. in Computer Science from the University of Missouri-Columbia. His research focuses on computational biology, bioinformatics, machine learning, and data science. His research group has developed novel computational and data-driven approaches for protein 3D modeling by combining cutting-edge machine learning and optimization algorithms, consistently achieving top-notch performance in worldwide Critical Assessment of Techniques for Protein Structure Prediction (CASP) experiments. He is a recipient of an NSF CAREER Award, an NIH Maximizing Investigators' Research Award (MIRA), and Auburn University College of Engineering’s Ginn Faculty Achievement Fellowship.