초록문의 abstract@kcsnet.or.kr

결제문의 member@kcsnet.or.kr

현재 가능한 작업은 아래와 같습니다.
  • 03월 08일 23시 이후 : 초록수정 불가능, 일정확인 및 검색만 가능

제127회 대한화학회 학술발표회 및 총회 Prediction of Protein Structure and Interaction by Deep Learning

2021년 2월 8일 13시 31분 18초
PHYS1-2 이곳을 클릭하시면 발표코드에 대한 설명을 보실 수 있습니다.
목 11시 : 00분
Physical Chemistry - Recent Research Trends in Biophysical Chemistry
저자 및
Chaok Seok
Division of Chemistry, Seoul National University, Korea
The history of protein structure prediction was driven by CASP competition, and recently DeepMind showed a monumental achievement in this field. One of the most interesting questions was that whether protein structures could be predicted by present biological information (protein sequences and structures) alone or not. For the last 26 years of CASP, bioinformatics approaches have been more successful in producing predictions than physics-based approaches, although some promising results were obtained by molecular dynamics simulations. In a related field of predicting protein-protein interactions represented by CAPRI and another field of predicting protein-ligand interactions, physics-based approaches have been more relevant. The absolute success of AlphaFold in CASP proved that protein structures can be predicted by exploiting present biological data at least for proteins forming stable well-defined tertiary structures. An important aspect of AlphaFold seems to be incorporation of physical insights and geometric elements into the architecture of neural network. This is because the current size of the protein structure database is not large enough compared to many other areas on which deep learning has shown great impacts. It remains to be seen how deep learning would change the field of predicting biomolecular interactions involving flexible and disordered molecules. In this talk, I will present a new deep learning approach that that attempts prediction of protein structures and interactions by neural net-based free energy function, which can be potentially applied to the cases involving structural flexibilities.