Alkin Kaz will present his General Exam "Ab Initio Reconstruction Ensembles for Resolving Heterogeneity in Cryo-EM" on Friday, May 8, 2026 at 11:00 AM in Friend 009. Committee Members: Ellen Zhong (advisor), Ben Raphael, Jia Deng Abstract: Biomolecular structure elucidation through cryo-electron microscopy relies heavily on computational pose-free (ab initio) scene reconstruction methods. Increasing capabilities of these methods enable successful capture of structures in increasingly relaxed experimental settings, where multiple different molecules present in the samples can be identified and reconstructed individually in data processing time. For the datasets which lie outside of the present capability frontiers, researchers resort to ad hoc and subjective exploration strategies such as iterative filtering and consistency across replicate runs. In this work, we expand upon such strategies under a framework of “reconstruction ensembles” that can systematically reconstruct heterogeneous datasets by considering the ab initio methods as weak classifiers to be aggregated. Through the construction of a weighted class assignment similarity score between particles across the job ensemble, we demonstrate the utility of this approach by exceeding the state-of-the-art reconstruction capabilities on the heterogeneous benchmark datasets. Reading List: https://docs.google.com/document/d/1zJHxkur2R9xy02JhwpgtS_nDv5Pe4jG7YO7L7aJj... Everyone is invited to attend the talk, and those faculty wishing to remain for the oral exam following are welcome to do so.