HIT-2: Implementing machine learning algorithms to treat bound ions in biomolecules
Department
Information Technology
Document Type
Article
Publication Date
1-1-2023
Abstract
Electrostatic features are fundamental to protein functions and protein-protein interactions. Studying highly charged biomolecules is challenging given the heterogeneous distribution of the ionic cloud around such biomolecules. Here we report a new computational method, Hybridizing Ions Treatment-2 (HIT-2), which is used to model biomolecule-bound ions using the implicit solvation model. By modeling ions, HIT-2 allows the user to calculate important electrostatic features of the biomolecules. HIT-2 applies an efficient algorithm to calculate the position of bound ions from molecular dynamics simulations. Modeling parameters were optimized by machine learning methods from thousands of datasets. The optimized parameters produced results with errors lower than 0.2 Å. The testing results on bound Ca2+ and Zn2+ in NAMD simulations also proved that HIT-2 can effectively identify bound ion types, numbers, and positions. Also, multiple tests performed on HIT-2 suggest the method can handle biomolecules that undergo remarkable conformational changes. HIT-2 can significantly improve electrostatic calculations for many problems in computational biophysics.
Journal Title
Computational and Structural Biotechnology Journal
Journal ISSN
2001-0370
Volume
21
First Page
1383
Last Page
1389
Digital Object Identifier (DOI)
10.1016/j.csbj.2023.02.013