The biological mechanism of a protein is determined by its three-dimensional (3D) structure, which is encoded in its one-dimensional (1D) amino acid sequence. Protein structure knowledge is utilised to assess their biological mechanisms, it can also aid in the discovery of new therapeutic interventions that can inhibit or activate the proteins to cure target diseases. Protein misfolding has been linked to a variety of diseases, including: type II diabetes and neurodegenerative disorders such as Alzheimer’s, Parkinson’s, and Huntington’s disease. Given the knowledge gap between a protein’s 1D string of amino acid sequence and its 3D structure, developing AI methods that can accurately predict 3D protein structures has tangible importance in assisting new drug discovery and the comprehension of protein-folding diseases.

An AI network can be used to predict the 3D shape of a protein; based on its amino acid sequence. A Deep Learning method is applied to predict the structure of proteins based on its sequence. The central component is a convolutional neural network that was trained on Protein Data Bank structures to predict the distances between every pair of residues in a protein sequence, yielding a probabilistic estimate of a 64 x 64 region of the distance map. These regions are then tiled together to generate distance predictions for the entire protein, which is then used to generate the protein structure that conforms to the distance predictions.

Another deep learning–based drug-target interaction prediction model has been recently developed; its purpose is to predict binding affinities. It is based on chemical sequences as well as amino acid sequences of a target protein but without its structural information. This approach can be used when it comes to identifying potent approved drugs; specifically those that might hinder the functions of the core proteins of SARS-CoV-2’s.