Towards prediction of multi-spin state charges of heme model by random forest regression
Abstract
The random forest regression (RFR) model was introduced to predict multiple spin state charges of heme model, which is important for molecular dynamic simulation of spin crossover phenomenon. In this work, A multiple spin state structure data set with 39368 structures of the simplified heme-oxygen binding model was built from the non-adiabatic dynamic simulation trajectories. The ESP charges of each atom were calculated and used as the real-valued response. The conformation adapted charge model (CAC) of three spin states was constructed by RFR using symmetry functions. The results show that our RFR can effectively predict the on the fly atomic charge varying with the conformation, and the atomic charge of different spin states in the same conformation as well, achieving the balance of accuracy and efficiency. The average mean absolute error of the predicted charges of each spin state is less than 0.02 e. The comparison studies on descriptors showed a maximum 0.06 e improvement in prediction of the charge of F e 2+ by using a manually selected 11 structural parameters. We hope that this model can not only provide variable parameters for developing the force field of multi-spin state, but also facilitate automation thus enable large-scale simulations of atomistic systems.
- Publication:
-
Frontiers in Chemistry
- Pub Date:
- March 2020
- DOI:
- 10.3389/fchem.2020.00162
- Bibcode:
- 2020FrCh....8..162Z
- Keywords:
-
- spin crossover;
- force field;
- machine learning;
- ESP charge;
- Heme model