Nuno M. Garrido

1-Octanol/water partition coefficients of n-Alkanes from molecular simulations of absolute solvation free energies

The 1-octanol/water partition coefficient is an important thermodynamic variable usually employed to understand and quantify the partitioning of solutes between aqueous and organic phases. It finds widespread use in many empirical correlations to evaluate the environmental fate of pollutants as well as in the design of pharmaceuticals. The experimental evaluation of 1-octanol/water partition coefficients is an expensive and time-consuming procedure, and thus, theoretical estimation methods are needed, particularly when a physical sample of the solute may not yet be available, such as in pharmaceutical screening. 1-Octanol/water partition coefficients can be obtained from Gibbs free energies of solvation of the solute in both the aqueous and the octanol phases. The accurate evaluation of free energy differences remains today a challenging problem in computational chemistry. In order to study the absolute solvation Gibbs free energies in 1-octanol, a solvent that can mimic many properties of important biological systems, free energy calculations for n-alkanes in the range C1-C8 were performed using molecular simulation techniques, following the thermodynamic integration approach. In the first part of this paper, we test different force fields by evaluating their performance in reproducing pure 1-octanol properties. It is concluded that all-atom force fields can provide good accuracy but at the cost of a higher computational time compared to that of the united-atom force fields. Recent versions of united-atom force fields, such as Gromos and TraPPE, provide satisfactory results and are, thus, useful alternatives to the more expensive all-atom models. In the second part of the paper, the Gibbs free energy of solvation in 1-octanol is calculated for several n-alkanes using three force fields to describe the solutes, namely Gromos, TraPPE, and OPLS-AA. Generally, the results obtained are in excellent agreement with the available experimental data and are of similar accuracy to commonly used QSPR models. Moreover, we have estimated the Gibbs free energy of hydration for the different compounds with the three force fields, reaching average deviations from experimental data of less than 0.2 kcal/mol for the case of the Gromos force field. Finally, we systematically compare different strategies to obtain the 1-octanol/water partition coefficient from the simulations. It is shown that a fully predictive method combining the Gromos force field in the aqueous phase and the OPLS-AA/TraPPE force field for the organic phase can give excellent predictions for n-alkanes up to C8 with an absolute average deviation of 0.1 log P units to the experimental data.

Using molecular simulation to predict solute solvation and partition coefficients in solvents of different polarity

A methodology is proposed for the prediction of the Gibbs energy of solvation (Δ(Solv)G) based on MD simulations. The methodology is then used to predict Δ(Solv)G of four solutes (namely propane, benzene, ethanol and acetone) in several solvents of different polarities (including n-hexane, n-hexadecane, ethylbenzene, 1-octanol, acetone and water) while testing the validity of the TraPPE force field parameters. Excellent agreement with experimental data is obtained, with average deviations of 0.2, 1.1, 0.8 and 1.2 kJ mol(-1), for the four solutes respectively. Subsequently, partition coefficients (log P) for forty different solute/solvent systems are predicted. The a priori knowledge of partition coefficient values is of high importance in chemical and pharmaceutical separation process design or as a measure of the increasingly important environmental fate. Here again, the agreement between experimental data and simulation predictions is excellent, with an absolute average deviation of 0.28 log P units. However, this deviation can be decreased down to 0.14 log P units, just by optimizing partial atomic charges of acetone in the water phase. Consequently, molecular simulation is proven to be a tool with strong physical basis able to predict log P with competitive accuracy when compared to the popular statistical methods with weak physical basis.

Development of a united-atom force field for 1-ethyl-3-methylimidazolium tetracyanoborate ionic liquid

Three united-atom (UA) force fields are presented for the ionic liquid 1-ethyl-3-methylimidazolium tetracyanoborate, abbreviated as [EMIM]+[B(CN)4]−. The atomistic charges were calculated based on the restrained electrostatic potential (RESP) of the isolated ions (abbreviated as force field 1, FF-1) and the ensemble averaged RESP (EA-RESP) method from the most stable ion pair configurations obtained by MP2/6-31G*+ calculations (abbreviated as FF-2 and FF-3). Non-electrostatic parameters for both ions were taken from the literature and Lennard–Jones parameters for the [B(CN)4]− anion were fitted in two different ways to reproduce the experimental liquid density. Molecular dynamics (MD) simulations were performed over a wide temperature range to identify the effect of the electrostatic and non-electrostatic potential on the liquid density and on transport properties such as self-diffusion coefficient and viscosity. Predicted liquid densities for the three parameter sets deviate less than 0.5% from experimental data. The molecular mobility with FF-2 and FF-3 using reduced charge sets is appreciably faster than that obtained with FF-1. FF-3 presents a refined non-electrostatic potential that leads to a notable improvement in both transport properties when compared to experimental data.

Predicting hydration Gibbs energies of alkyl-aromatics using molecular simulation: a comparison of current force fields and the development of a new parameter set for accurate solvation data

The Gibbs energy of hydration is an important quantity to understand the molecular behavior in aqueous systems at constant temperature and pressure. In this work we review the performance of some popular force fields, namely TraPPE, OPLS-AA and Gromos, in reproducing the experimental Gibbs energies of hydration of several alkyl-aromatic compounds--benzene, mono-, di- and tri-substituted alkylbenzenes--using molecular simulation techniques. In the second part of the paper, we report a new model that is able to improve such hydration energy predictions, based on Lennard Jones parameters from the recent TraPPE-EH force field and atomic partial charges obtained from natural population analysis of density functional theory calculations. We apply a scaling factor determined by fitting the experimental hydration energy of only two solutes, and then present a simple rule to generate atomic partial charges for different substituted alkyl-aromatics. This rule has the added advantages of eliminating the unnecessary assumption of fixed charge on every substituted carbon atom and providing a simple guideline for extrapolating the charge assignment to any multi-substituted alkyl-aromatic molecule. The point charges derived here yield excellent predictions of experimental Gibbs energies of hydration, with an overall absolute average deviation of less than 0.6 kJ mol(-1). This new parameter set can also give good predictive performance for other thermodynamic properties and liquid structural information.

Predicting hydrophobic solvation by molecular simulation: 1. testing united-atom alkane models

We present a systematic test of the performance of three popular united‐atom force fields—OPLS‐UA, GROMOS and TraPPE—at predicting hydrophobic solvation, more precisely at describing the solvation of alkanes in alkanes. Gibbs free energies of solvation were calculated for 52 solute/solvent pairs from Molecular Dynamics simulations and thermodynamic integration making use of the IBERCIVIS volunteer computing platform. Our results show that all force fields yield good predictions when both solute and solvent are small linear or branched alkanes (up to pentane). However, as the size of the alkanes increases, all models tend to increasingly deviate from experimental data in a systematic fashion. Furthermore, our results confirm that specific interaction parameters for cyclic alkanes in the united‐atom representation are required to account for the additional excluded volume within the ring. Overall, the TraPPE model performs best for all alkanes, but systematically underpredicts the magnitude of solvation free energies by about 6% (RMSD of 1.2 kJ/mol). Conversely, both GROMOS and OPLS‐UA systematically overpredict solvation free energies (by ∼13% and 15%, respectively). The systematic trends suggest that all models can be improved by a slight adjustment of their Lennard‐Jones parameters.