|
|
| Type |
Poster Presentation |
| Area |
Physical Chemistry |
| Room No. |
Event Hall |
| Time |
4월 20일 (금요일) 11:00~12:30 |
| Code |
PHYS.P-188 |
| Subject |
Using Contact Map and Solvent Accessible Surface Area on Deep Learning Improves Protein Free Energy Prediction |
| Authors |
Woohyun Kim, Sihyun Ham1,*
Department of Chemistry, Sookmyung Womens' University, Korea
1Department of Chemistry, Sookmyung Women's University, Korea
|
| Abstract |
Calculating the solvation free energy of a protein would require substantial computation and takes large resources. In this study, we have developed an algorithm that rapidly predict the solvation free energies as well as the internal potential energies of proteins. The algorithm is based on deep neural network (DNN) using the contact maps and residual solvent accessible surface areas (SASA) as input features. The contact map represents the distances between all residue pairs and the residual SASA was divided into polar and nonpolar contributions. We found that the accuracy of the prediction increases by using contact map and residual SASA together compared to using only contact map as input data. We conclude that the thermodynamic quantities of proteins can be predicted accurately and quickly with the algorithm proposed. |
| E-mail |
woosa7@gmail.com |
|
121st General Meeting of the KCS