Group Contribution and Machine Learning Approaches to Predict Abraham Solute Parameters, Solvation Free Energy, and Solvation Enthalpy

We present a group contribution method (SoluteGC) and a machine learning model (SoluteML) to predict the Abraham solute parameters, as well as a machine learning model (DirectML) to predict solvation free energy and enthalpy at 298 K. The proposed group contribution method uses atom-centered functional groups with corrections for ring and polycyclic strain whilst the machine learning models adopt a directed message passing neural network. The solute parameters predicted from SoluteGC and SoluteML are used to calculate solvation energy and enthalpy via linear free energy relationships. Extensive data sets containing 8366 solute parameters, 20253 solvation free energies, and 6322 solvation enthalpies are compiled in this work to train the models. The three models are each evaluated on the same test sets using both random and substructure-based solute splits for solvation energy and enthalpy predictions. The results show that the DirectML model is superior to the SoluteML and SoluteGC models for both predictions and can provide accuracy comparable to that of advanced quantum chemistry methods. Yet, even though the DirectML model performs better in general, all three models are useful for various purposes. Uncertain predicted values can be identified by comparing the 3 models, and when the 3 models are combined together, they can provide even more accurate predictions than any one of them individually. Finally, we present our compiled solute parameter, solvation energy, and solvation enthalpy databases (SoluteDB, dGsolvDBx, dHsolvDB) and provide public access to our final prediction models through a simple web-based tool, software package, and source code.

Estimating Activity Coefficients of Target Components in Poorly Specified Mixtures with NMR Spectroscopy and COSMO-RS

Poorly specified mixtures are common in process engineering, especially in bioprocess engineering. The properties of such mixtures of unknown composition cannot be described using conventional thermodynamic models. The NEAT method, which has recently been developed in our group, enables the calculation of activity coefficients of known target components in such poorly specified mixtures. In NEAT, the group composition of the mixture is determined by NMR spectroscopy and a thermodynamic group contribution method is used for calculating the activity coefficients. In all previous studies with NEAT, the UNIFAC group contribution method was used. In the present work, we demonstrate that NEAT can also be applied with another important method for predicting activity coefficients: COSMO-RS. COSMO-RS (OL) developed in Oldenburg together with its group contribution version GC-COSMO-RS (OL) is used here. The new version of NEAT was successfully tested. For a variety of aqueous mixtures excellent agreement of the NEAT predictions, for which only information on the target component was used, with results that were obtained using the full knowledge on the composition of the mixture was found. The results demonstrate the generic nature of the idea of NEAT and the broad applicability of the method.

Group Contribution and Machine Learning Approaches to Predict Abraham Solute Parameters, Solvation Free Energy, and Solvation Enthalpy

We present a group contribution method (SoluteGC) and a machine learning model (SoluteML) to predict the Abraham solute parameters, as well as a machine learning model (DirectML) to predict solvation free energy and enthalpy at 298 K. The proposed group contribution method uses atom-centered functional groups with corrections for ring and polycyclic strain whilst the machine learning models adopt a directed message passing neural network. The solute parameters predicted from SoluteGC and SoluteML are used to calculate solvation energy and enthalpy via linear free energy relationships. Extensive data sets containing 8366 solute parameters, 20253 solvation free energies, and 6322 solvation enthalpies are compiled in this work to train the models. The three models are each evaluated on the same test sets using both random and substructure-based solute splits for solvation energy and enthalpy predictions. The results show that the DirectML model is superior to the SoluteML and SoluteGC models for both predictions and can provide accuracy comparable to that of advanced quantum chemistry methods. Yet, even though the DirectML model performs better in general, all three models are useful for various purposes. Uncertain predicted values can be identified by comparing the 3 models, and when the 3 models are combined together, they can provide even more accurate predictions than any one of them individually. Finally, we present our compiled solute parameter, solvation energy, and solvation enthalpy databases (SoluteDB, dGsolvDBx, dHsolvDB) and provide public access to our final prediction models through a simple web-based tool, software package, and source code.

Using the group contribution method and molecular dynamics to predict the glass transition temperatures and mechanical properties of poly-(p-phenylenediamine-alt-2, 6-diformyl multiphenyl)

This article reports the glass transition temperatures of poly-( p-phenylenediamine-alt-2,6-diformyl multiphenyl) predicted by both the group contribution method and the molecular dynamics simulations. The related modeling method and the degree of polymerization, density, specific volume, radius of volume, radius of rotation, and non-bonding energy terms with temperature are analyzed in depth. The bulk modulus, shear modulus, compressibility, Young’s modulus, and Poisson’s ratio of poly-( p-phenylenediamine-alt-2,6-diformyl multiphenyl) at room temperature are simulated by molecular dynamics. The results show that the simulated glass transition temperatures of poly-( p-phenylenediamine-alt-2,6-diformyl multiphenyl) are greater than 480 K, which indicates that poly-( p-phenylenediamine-alt-2,6-diformyl multiphenyl) can be expected to be used as a high-temperature-resistant material. As the number of rigid benzene rings on the molecular side chain increases, the glass transition temperature decreases, with an average decrease of 10 K for each additional benzene ring. The free volume theory can explain the glass transitions of poly-( p-phenylenediamine-alt-2,6-diformyl multiphenyl). The modulus and density of poly-( p-phenylenediamine-alt-2,6-diformyl multiphenyl) change accordingly with an increase of rigid benzene rings on the side chain, probably due to the fact that the flexibility of the polymers changes accordingly as the number of benzene rings on the side chain increases.

NEAT-NMR Spectroscopy for the Estimation of Activity Coefficients of Target Components in Poorly Specified Mixtures

Mixtures of which the composition is not fully known are important in many fields of engineering and science, for example, in biotechnology. Owing to the lack of information on the composition, such mixtures cannot be described with common thermodynamic models. In the present work, a method is described with which this obstacle can be overcome for an important class of problems. The method enables the estimation of the activity coefficients of target components in poorly specified mixtures and is based on a combination of NMR spectroscopy with a thermodynamic group contribution method. It is therefore called the NEAT method (NMR spectroscopy for the Estimation of Activity coefficients of Target components in poorly specified mixtures). In NEAT, NMR spectroscopy is used to obtain information on the concentrations of chemical groups in the mixture. The elucidation of the speciation is not required, only the target component has to be known. Modified UNIFAC (Dortmund) is applied in the present work as group contribution method, but NEAT can be extended to any other group contribution method. NEAT was introduced recently by our group in a short communication, but only the basic ideas were presented. In the present work, NEAT is described in full detail. Different options of using NEAT are discussed, and examples for the application of the method are given. They include a variety of aqueous and nonaqueous mixtures. The results show very good agreement of the activity coefficients that are predicted by NEAT with the corresponding results for the fully specified mixtures.