Progress in Determination of Protein Spatial Structure Based on Machine Learning

Introduction. The task of determining the spatial structure proteins is one most important unsolved problems mankind. Life on planet Earth called protein, because protein molecules are drivers life processes in living organisms. Proteins make up about 80% dry mass cell and coordinate metabolism. functions defined by its structure. results recent competitions methods for structures have shown significant progress this area. One research groups presented AlphaFold 2 method, accuracy which reached experimental methods. Purpose article. aim work to consider analyze basic principles software package proteins. Results. We main stages process recognizing a using program complex. corresponding include: search homologous based multiple alignment methods, construction protein-specific differentiated potential artificial neural networks energy optimization gradient descent limited sampling. discuss how combination various bioinformatics techniques powered data from open sources can lead improvements prediction. Special attention paid use building smooth following minimization constructed potential. Conclusions. number information genetic banks allows us solving extremely protein. Keywords: structure, Machine Learning, AlphaFold.