Lessons Learned from Multiobjective Automatic Optimizations of Classical Three-Site Rigid Water Models Using Microscopic and Macroscopic Target Experimental Observables

Abstract

The development of accurate water models is of primary importance for molecular simulations. Despite their intrinsic approximations, three-site rigid water models are still ubiquitously used to simulate a variety of molecular systems. Automatic optimization approaches have been recently used to iteratively refine three-site water models to fit macroscopic (average) thermodynamic properties, providing state-of-the-art three-site models that still present some deviations from the liquid water properties. Here, we show the results obtained by automatically optimizing three-site rigid water models to fit a combination of microscopic and macroscopic experimental observables. We use Swarm-CG , a multiobjective particle-swarm-optimization algorithm, for training the models to reproduce the experimental radial distribution functions of liquid water at various temperatures (rich in microscopic-level information on, e.g., the local orientation and interactions of the water molecules). We systematically analyze the agreement of these models with experimental observables and the effect of adding macroscopic information to the training set. Our results demonstrate how adding microscopic-rich information in the training of water models allows one to achieve state-of-the-art accuracy in an efficient way. Limitations in the approach and in the approximated description of water in these three-site models are also discussed, providing a demonstrative case useful for the optimization of approximated molecular models, in general.

Conformational Preferences at the Glycosidic Linkage of Saccharides in Solution as Deduced from NMR Experiments and MD Simulations: Comparison to Crystal Structures

“…The 3 J CH (ψ) values differ between the three disaccharides, which is consistent with different averages of the ψ torsion angle in each disaccharide. However, for 1 and 2 the 3 J CH (ψ) values are overestimated from the MD simulations, whereas for 3 it is underestimated, which calls for further evaluation of Karplus-type relationships, on the one hand, and glycan-specific force fields as well as water models [74][75][76][77] used for MD simulations, on the other. Moreover, in combining an additive empirical carbohydrate force field with a certain water model, care has to be taken in optimizing partial atomic charges and the Lennard-Jones parameters of different interactions, referred to as the interaction triad, i. e., intermolecular water-water, intermolecular water-carbohydrate and intramolecular carbohydrate-carbohydrate interactions.…”

Building Water Models Compatible with Charge Scaling Molecular Dynamics

“…The first simulations involving water date back to the early days of FFMD . Consequently, the development of empirical potentials for water has been a recurrent topic in the past decades (e.g, TIPS, SPC, TIP3P, SPC/E, and TIP4P models) and is far from settled (e.g., the more recent TIP4P/2005, TIP4P-FB, and four-site OPC (OPC4) models). Aqueous solutions have proven to be difficult systems to describe accurately and are thus an active area of research .…”

“…Aqueous solutions have proven to be difficult systems to describe accurately and are thus an active area of research . Even pure water behavior is not easy to model such that it accurately covers the full range of biologically relevant thermodynamic conditions …”

“… 12 Even pure water behavior is not easy to model such that it accurately covers the full range of biologically relevant thermodynamic conditions. 8 …”