Online, Once a Month. Wednesday at 8am PT / 11am ET/ 4pm BST / 5pm CET / 11pm China.
Registration has changed. For zoom links and announcements, please register to our mailing list. This is a read-only mailing list (users cannot post). Please use your institution/ company email – we approve registrations manually and this helps us know who you are . You may receive an email asking you to confirm your registration, it is important that you click the link, otherwise we will not see your request.
Organizers: Joakim Andén, Dorit Hanein, Roy R. Lederman and Steven J. Ludtke
All Talks – Past talks – Upcoming Talks
Next Talk
Jianlin (Jack) Cheng
William and Nancy Thompson Missouri Distinguished Professor in Electrical Engineering and Computer Science
University of Missouri
Date and time: November 4th, 2026 Wednesday at 8am PST / 11am EST / 4pm BST / 5pm CET / 11pm China
For the zoom links, please join the One World Cryo-EM mailing list.
AI for Cryo-EM Data Analysis: From Particle Picking to Atomic Model Building
Cryo-electron microscopy (cryo-EM) has emerged as a powerful technology for determining the structures of proteins and large molecular complexes, including challenging assemblies that are difficult to study using other structural biology approaches. However, extracting accurate structural information from cryo-EM data remains challenging and labor-intensive, particularly for noisy micrographs, heterogeneous density maps, and large or complex protein assemblies. Artificial intelligence (AI) offers new opportunities to automate and improve these analyses. In this talk, I will present several AI-based methods and tools developed in my lab for key stages of cryo-EM data analysis, including particle picking, density map enhancement, and atomic model building. I will first introduce CryoTransformer and CryoSegNet, two AI-based methods for automated and accurate particle picking from cryo-EM micrographs. I will then discuss CryoTEN, a transformer-based approach for enhancing cryo-EM density maps and improving their interpretability. Finally, I will present Cryo2Struct, a deep learning approach for de novo atomic model building, and MICA, a multimodal AI framework that integrates cryo-EM density maps with AlphaFold3 predictions to build accurate protein structures. I will highlight benchmark results and examples demonstrating the potential of these methods and tools to make cryo-EM structure determination more accurate, automated, and robust.
Last Talk
Minhuan Li
Flatiron Research Fellow
Flatiron Institute
Date and time: October 7th, 2026 Wednesday at 8am PST / 11am EST / 4pm BST / 5pm CET / 11pm China
For the zoom links, please join the One World Cryo-EM mailing list.
From Prediction to Scientific Inference: Grounding Generative Models in Biological Experiments
Biological measurements and predictive models are both scaling rapidly, yet the interface between them remains a bottleneck. In structural biology, deposited structures have enabled major advances in AI but capture only part of the underlying experimental evidence. Connecting models more directly to measurements could better exploit information about conformational populations, experimental context, and uncertainty.
This talk presents experiment-grounded generative inference: combining learned priors with measurements to infer sample-specific biological states. I will develop these ideas primarily in structural biology, where the states of interest are biomolecular conformations. I will focus on ROCKET, which uses AlphaFold as a learned prior for inference from X-ray crystallography and cryo-EM measurements, and EmbedOpt, which steers conditional protein diffusion models by optimizing their conditioning embeddings at inference time. Both approaches use pretrained models as structured search spaces, allowing experimental evidence to guide inference beyond a model’s default prediction. I will also briefly discuss GOTO, an optimal-transport-based objective that provides informative optimization signals for joint inference of molecular states and cryo-EM imaging parameters.
I will close with an outlook on MIMIC and possible routes for making measurements first-class inputs and training signals in multimodal biomolecular models. The broader goal is to use learned priors to interpret experimental evidence—and ultimately use that evidence to improve the priors themselves.
References:
Golkar, Li, et al. — MIMIC (arXiv, 2026)
Fadini, Li et al. — AlphaFold as a prior (Nature Methods, 2026)
Li et al. — EmbedOpt (arXiv, 2026)
Woollard, Li et al. — GOTO (bioRxiv, 2025)