The Multiscale Fragment Method With Cutting the Chemical Bonds of System: <scp>MD</scp> Simulations of Trypsin Inhibitor Aqueous Solution by <scp>ABEEM</scp> Polarizable Force Field

ABSTRACT Long‐time molecular dynamics (MD) simulations of large systems with solvent are a major challenge for theoretical approaches. The popular method is the fragment method, and the challenge of which lies in how to deal with the chemical bonds that are cut and the setting of charges in each subdivision. Dividing the periodic box into subdivisions with their surroundings (DBSS) method has been developed to solve the aforementioned two problems. The DBSS method divides the periodic box or unit of the system into subdivisions according to the atomic coordinates, and solves the charge distribution of each subdivision separately. Partial charges of atoms at the boundary of the subdivision are crucial for computational accuracy. If the cut chemical bond is located in a charged residue, the entire residue is assigned to the subdivision that contains the key atom of the residue. If the cut chemical bond is located in a neutral residue, the partial charge of the bond site is assigned equally between two subdivisions. The total charge of each subdivision is set as the sum of the charges of the charged residues. For non‐bonding interactions at the boundary, the DBSS method extends the 4 Å surroundings for each subdivision, which solves the problem of neglecting long‐range electrostatic interactions in conventional fragment methods. The ABEEM‐DBSS method was applied to perform MD simulations on five trypsin/inhibitor complexes. The ABEEM‐7P water model is used for the 3 Å solute surrounding the complexes, and the TIP‐3P water model is applied to the distant aqueous solution. The computational efficiency is optimal when each subdivision contains 200–300 atoms. If the system is divided into M subdivisions, the computation time is reduced by a factor of M 3 approximately. The size of the simulated system is 100 times larger than that of the original ABEEM method. Compared to the OPLS force field, the deviations of structures are smaller. The predicted sequence of binding energies is consistent with the inhibition constants by photometric assay. It is evident that ABEEM‐DBSS is a reliable tool for fast and accurate simulations of macromolecular systems.