One-shot Cryo-EM Complex Structure Determination with High Accuracy and Ultra-fast Speed.

While cryo-electron microscopy (Cryo-EM) yields high-resolution density maps for complex structures, accurate determination of the corresponding three-dimensional atomic structures still necessitates significant expertise and labor-intensive manual interpretation. Recently, AI-based methods have emerged to streamline this process in the biological community; however, several challenges persist. First, existing methods typically require multi-stage training and inference, causing inefficiencies and inconsistency between stages. Second, these approaches often encounter bias in aligning predicted atomic coordinates with sequence. Researchers have utilized Hidden Markov Model(HMM) or Traveling Salesman Problem (TSP) algorithms to explore the sequence space, which incurs substantial computational costs. Lastly, due to limitations of available datasets, prior works struggle to generalize effectively to complicated and unseen test data, a common problem in machine learning. In response to these challenges, we introduce End-to-End and Efficient CryoFold, or E3-CryoFold for short, a deep learning method that enables end-to-end training and one-shot inference. E3-CryoFold employs both 3D and sequence Transformers to extract features from density maps and sequences, using cross-attention modules to integrate the two modalities. Additionally, it utilizes an equivariant graph neural network to construct the atomic structure based on the extracted features. Importantly, E3-CryoFold incorporates a pretraining stage, during which models are trained on simulated density maps derived from Protein Data Bank (PDB) structures. Empirical results demonstrate that E3-CryoFold improves the average TM-score of the generated structures by 400\% as compared to Cryo2Struct and achieves this huge improvement using merely 1/1000 of the inference time as required by Cryo2Struct. Thus, E3-CryoFold represents a robust, streamlined, and cohesive paradigm for Cryo-EM structure determination.