A Study on the End-to-End AlphaFold 3D Protein Structure Prediction Method

Nearly all of life's functions depend on proteins, which are fundamental to life. One of the difficulties explored in biological research is the interaction and folding of polypeptide chains into three-dimensional structures through chains of amino acids, a process known as protein folding. In this essay, we begin by outlining the problem's history and the previous work that has been done. The open source protein database PDB is used in a deep learning method to create a model for predicting protein structure based on an end-to-end network. The proposed multiple parallel attention model built on AlphaFold can be regarded as being superior as a result of the experiment. We first introduce the background knowledge related to protein prediction, then we give a detailed introduction to the protein database PDB. Next we introduce the neural network principle and Alphafold network results related to it. Finally we introduce the core of this paper that is AlphaFold based multiple parallel attention end-to-end model, and then we compare the effect of five different model structures AlphaFold model, AlphaFold + Attention+Dense, AlphaFold + Attention + LSTM, AlphaFold + Attention + GRU and AlphaFold multiple parallel attention, respectively. And the final results show that AlphaFold-based multiple parallel attention end-to-end model gives the best results.