[20a-P03-1~29] 12.6 Nanobiotechnology

[20a-P03-28] Calculation of quantitative interaction energy from low-cost FMO calculations by machine learning – part 1

〇Ryohei Yoshine¹, Hideo Doi¹, Sota Matsuoka¹, Koji Okuwaki^1,2, Yuji Mochizuki^1,3 (1.Rikkyo Univ., 2.JSOL Corp., 3.Univ. Tokyo.)

Keywords:

Machine Learning,FMO,IFIE

In recent years, dynamic and statistical interaction analyses have been performed in conjunction with MD and FMO, taking into account fluctuations. However, the high computational cost is a challenging problem. Therefore, we are now working on a project to reduce the computational cost by using machine learning to predict the intermolecular interaction energy (IFIE) with high-cost basis functions based on the results of FMO calculations with low-cost basis functions. On the presentation day, we will report on the comparison of calculated and predicted values and the change in prediction accuracy as descriptors are added, using Chignolin as an example.