Cryoelectron microscope atomic model structure building method and system based on deep learning and application

Abstract

The invention discloses a cryoelectron microscope atomic model structure building method and system based on deep learning and application. The method comprises the steps of: 1, obtaining a cryoelectron microscope density map data set, and carrying out model training and model testing; 2, inputting a cryoelectron microscope density map and a corresponding amino acid sequence; and 3, carrying out feature coding and extraction on the cryoelectron microscope density map and the corresponding amino acid sequence, and building an atomic structure model. According to the measurement method provided by the invention, the generated amino acid atom model has structural biological characteristics, the structural biological rationality of the predicted amino acid atom model is ensured, accurate prediction of the end-to-end all-differentiable amino acid internal atom structure is finally realized, certain superiority is achieved, and the atomic model effect predicted by a plurality of tests is verified. In addition, the improvement effect in model building in middle and low resolution is also very obvious.

Description

technical field [0001] The invention belongs to the technical field of structural biology, and in particular relates to a cryo-electron microscope atomic model structure building method, system and application, in particular to a deep learning-based cryo-electron microscope atomic model structure building method, system and application. Background technique [0002] In traditional machine learning, the execution process of a task often consists of many modules. It is generally divided into multiple independent steps such as data preprocessing, feature extraction, model training, and result post-processing, which are integrated and executed to realize automatic operation. However, the quality of the result of each step will affect each subsequent step, thus affecting the quality of the final result to a certain extent. In addition, for supervised learning, the label used for supervision will contribute to each link, but because the intermediate steps are too independent, a l...

AI Technical Summary

Problems solved by technology

However, in the density map with lower resolution, due to the missing or low side chain density, the position regressed by the heat map based on the density feature is biased towards the main chain

Images