“DeepMASC: Automated Mask Creation and Class Selection in Single Particle Analysis Empowered by Deep Protein Probability Measures of Cryo-EM Maps”
_
Date and Time
4:00 PM – 5:30 PM Monday, June 16th, 2025
Location
Hybrid (Zoom and KEK CryoEM building)
Speakers:
Mr. Han Zhu, M.S.
Department of Computer Science, Purdue University, US.
Abstract
We have developed DeepMASC (Deep Masking and Auto-Selection for Classifications), a novel approach that eliminates human intervention from two critical steps in the cryo-EM single particle analysis (SPA) computational workflow: 3D class selection and 3D reference mask generation. Our method leverages deep learning-based protein probability measures to automatically identify optimal structural classes and generate appropriate masks for further refinement. DeepMASC dramatically accelerates the SPA workflow by eliminating time-consuming manual interventions that often create significant delays. Traditional workflows require expert decisions at critical junctures, creating bottlenecks that can extend processing time from hours to days. By automating these decision points, DeepMASC enables continuous, uninterrupted processing without sacrificing quality.
To address broader computational challenges in cryo-EM analysis, we also have developed the Kihara Lab EMSuite Server, a free and comprehensive web platform offering 13 algorithms for cryo-EM analysis. The server spans the full spectrum of needs, from secondary structure detection at medium resolution to atomic modeling at high resolution for proteins, nucleic acids, and their complexes. This unified interface makes sophisticated cryo-EM analysis accessible to structural biologists regardless of computational expertise, democratizing access to cutting-edge methods without requiring specialized hardware or software installation.
Profile
2018-2022 B.S., Department of Computer Science, Purdue University, US.
2022-2023 M.S., Department of Computer Science, Purdue University, US.
2023-Present Ph.D. program, Department of Computer Science, Purdue University, US
Research Focus
Computational Structural Biology, ML-Based Cryo-EM Modeling and Analysis