Abstract:The ff15ipq protein force field is a fixed charge model built by automated tools based on the two charge sets of the implicitly polarized charge method: one set (appropriate for vacuum) for deriving bonded parameters and the other (appropriate for aqueous solution) for running simulations. The duality is intended to treat water-induced electronic polarization with an understanding that fitting data for bonded parameters will come from quantum mechanical calculations in the gas phase. In this study, we compare ff15ipq to two alternatives produced with the same fitting software and a further expanded data set but following more conventional methods for tailoring bonded parameters (harmonic angle terms and torsion potentials) to the charge model. First, ff15ipq-Qsolv derives bonded parameters in the context of the ff15ipq solution phase charge set. Second, ff15ipq-Vac takes ff15ipq's bonded parameters and runs simulations with the vacuum phase charge set used to derive those parameters. The IPolQ charge model and associated protocol for deriving bonded parameters are shown to be an incremental improvement over protocols that do not account for the material phases of each source of their fitting data. Both force fields incorporating the polarized charge set depict stable globular proteins and have varying degrees of success modeling the metastability of short (5-19 residues) peptides. In this particular case, ff15ipq-Qsolv increases stability in a number of α-helices, correctly obtaining 70% helical character in the K19 system at 275 K and showing appropriately diminishing content up to 325 K, but overestimating the helical fraction of AAQAA3 by 50% or more, forming long-lived α-helices in simulations of a β-hairpin, and increasing the likelihood that the disordered p53 N-terminal peptide will also form a helix. This may indicate a systematic bias imparted by the ff15ipq-Qsolv parameter development strategy, which has the hallmarks of strategies used to develop other popular force fields, and may explain some of the need for manual corrections in this force fields' evolution. In contrast, ff15ipq-Vac incorrectly depicts globular protein unfolding in numerous systems tested, including Trp cage, villin, lysozyme, and GB3, and does not perform any better than ff15ipq or ff15ipq-Qsolv in tests on short peptides. We analyze the free energy surfaces of individual amino acid dipeptides and the electrostatic potential energy surfaces of each charge model to explain the differences.

Likely Overstabilization of Charge-Charge Interactions in CHARMM36m(w): A Case for a99SB-disp Water

Development of an Effective Polarizable Bond Method for Biomolecular Simulation.

A quantum mechanical computational method for modeling electrostatic and solvation effects of protein

Electronic Polarization Stabilizes Tertiary Structure Prediction of HP-36.

Balancing Group I Monoatomic Ion-Polar Compound Interactions for Condensed Phase Simulation in the Polarizable Drude Force Field

From bonds to interactions: comprehensive molecular characterization via polarizable bond-dipole approach

Building Water Models Compatible with Charge Scaling Molecular Dynamics

Accounting for electronic polarization in nonpolarizable force fields

Effective Inclusion of Electronic Polarization Improves the Description of Electrostatic Interactions: The prosECCo75 Biomolecular Force Field

Polarizable Force Field of Intrinsically Disordered Proteins with CMAP and Reweighting Optimization

CHARMM36m: an improved force field for folded and intrinsically disordered proteins

Stochastic Constrained Extended System Dynamics for Solving Charge Equilibration Models

Transferable Ion Force Fields in Water from a Simultaneous Optimization of Ion Solvation and Ion–Ion Interaction

Links between the charge model and bonded parameter force constants in biomolecular force fields

Systematic improvement of empirical energy functions in the era of machine learning

CHARMM-GUI Input Generator for NAMD, GROMACS, AMBER, OpenMM, and CHARMM/OpenMM Simulations Using the CHARMM36 Additive Force Field

Improved Parameterization of Amine-Carboxyate and Amine-Phosphate Interactions for Molecular Dynamics Simulations Using the CHARMM and AMBER Force Fields

Polarizable Mean-Field Model of Water for Biological Simulations with Amber and Charmm force fields

Significantly Improved Protein Folding Thermodynamics Using a Dispersion-Corrected Water Model and a New Residue-Specific Force Field

Polarizable Water Model with Ab Initio Neural Network Dynamic Charges and Spontaneous Charge Transfer