Searching the Molecular Space using Neural Network Energy Models
발표자
()
초록
내용
Protein structure prediction has become highly accurate due to the recent advances in deep learning techniques, and new methods for predicting biomolecular interactions and designing functional molecules have been reported to utilize such advances. However, recent structure prediction methods still fail to predict structures and interactions lacking evolutionary information. We have been developing structure prediction and molecular design methods to overcome such limitations. Traditional energy-based structure modeling methods that we have developed for the last twenty years are available on GalaxyWEB (https://galaxy.seoklab.org), and we are now using recently developed co-evolution-based methods in combination with traditional methods and developing new neural net energy models for more accurate structure prediction and molecular design. Applications in this avenue to the prediction of protein structures, protein-protein interactions, and protein-ligand interactions will be presented.