Novel Artificial Intelligence-Based Approaches for Ab Initio Structure Determination and Atomic Model Building for Cryo-Electron Microscopy

Abstract

Single particle cryo-electron microscopy (cryo-EM) has emerged as the prevailing method for near-atomic structure determination, shedding light on the important molecular mechanisms of biological macromolecules. However, the inherent dynamics and structural variability of biological complexes coupled with the large number of experimental images generated by a cryo-EM experiment make data processing nontrivial. In particular, ab initio reconstruction and atomic model building remain major bottlenecks that demand substantial computational resources and manual intervention. Approaches utilizing recent innovations in artificial intelligence (AI) technology, particularly deep learning, have the potential to overcome the limitations that cannot be adequately addressed by traditional image processing approaches. Here, we review newly proposed AI-based methods for ab initio volume generation, heterogeneous 3D reconstruction, and atomic model building. We highlight the advancements made by the implementation of AI methods, as well as discuss remaining limitations and areas for future development.

The advent of preventive high-resolution structural histopathology by artificial-intelligence-powered cryogenic electron tomography

“…These types of feats have been possible and will become increasingly common thanks to AI-assisted acceleration and improvement in various workflow steps. These include specimen screening before data collection (Bouvette et al, 2022;Cheng et al, 2023), micrograph denoising (Tegunov and Cramer, 2019;Bepler et al, 2020), structure reconstruction (Giri et al, 2023), andpostprocessing (Sanchez-Garcia et al, 2021), with particular efforts having been concentrated in particle picking (Wang et al, 2016;Zhu et al, 2017;Sanchez-Garcia et al, 2018;George et al, 2021) and model building (He et al, 2022;DiIorio and Kulczyk, 2023;Giri et al, 2023). Machine learning approaches have sped up cryoEM SPA structure determination to the point that, for many specimens, including those exhibiting compositional and conformational heterogeneity (Zhong et al, 2021), multiple structures at near-atomic resolution can be derived in a few days, sometimes even from a single imaging session and processing the corresponding data using a single workstation equipped with GPU acceleration (Kimanius et al, 2016).…”

The advent of preventive high-resolution structural histopathology by artificial-intelligence-powered cryogenic electron tomography

“…These types of feats have been possible and will become increasingly common thanks to AI-assisted acceleration and improvement in various workflow steps. These include specimen screening before data collection ( Bouvette et al, 2022 ; Cheng et al, 2023 ), micrograph denoising ( Tegunov and Cramer, 2019 ; Bepler et al, 2020 ), structure reconstruction ( Giri et al, 2023 ), and postprocessing ( Sanchez-Garcia et al, 2021 ), with recent efforts having been particularly concentrated on particle picking ( Wang et al, 2016 ; Zhu et al, 2017 ; Sanchez-Garcia et al, 2018 ; George et al, 2021 ) and model building ( He et al, 2022 ; DiIorio and Kulczyk, 2023 ; Giri et al, 2023 ). Machine learning approaches have sped up cryoEM SPA structure determination to the point that, for many specimens, including those exhibiting compositional and conformational heterogeneity ( Zhong et al, 2021 ), multiple structures at near-atomic resolution can be derived in a few days, sometimes even from a single imaging session and processing the corresponding data using a single workstation equipped with GPU acceleration ( Kimanius et al, 2016 ).…”

Polymerizing laminins in development, health, and disease

“…Recent advances in cryo-EM provide unprecedented insight into structures of dynamic macromolecular complexes at atomic resolution ( 151 , 152 , 153 , 154 , 155 ). We previously employed cryo-EM to determine a 3.7 Å structure of the laminin polymer node containing the N-terminal 56 kDa α1, 64 kDa β1, and 52 kDa γ1 fragments ( Fig.…”