Toward Predicting Solubility of Arbitrary Solutes in Arbitrary Solvents: Prediction of Density and Refractive Index...

Area of Science:

• Physical Chemistry
• Computational Chemistry
• Chemical Engineering

Background:

• Density and refractive index (nD) are critical for organic process development, aiding solvent selection, mixture formulation, and composition assessment.
• Accurate prediction of these properties is valuable for molecular mechanics force field development and predicting solvation free energy and solubility.
• These properties are linked to molecular van der Waals (VDW) energy.

Purpose of the Study:

• To develop robust machine learning (ML) models for predicting the density and refractive index (nD) of organic compounds.
• To identify key molecular descriptors contributing to model accuracy.
• To investigate the impact of temperature on prediction models and improve accuracy for outlier data points.

Main Methods:

• Gathered molecular characteristic data for ~5000 compounds (density) and ~4000 compounds (nD).
• Generated General AMBER Force Field (GAFF) and RDKit descriptors for training ML models.
• Employed various ML algorithms, optimized best-performing models, and performed global sensitivity and Shapley analyses.
• Investigated temperature effects and used molecular dynamics (MD) simulations for density outliers.

Main Results:

• Both GAFF and RDKit descriptors generated robust models with low errors (density APE: 2.67-3.15%, nD APE: 0.53%).
• RDKit descriptors showed slightly superior performance for both properties.
• Sensitivity analyses identified key features contributing to model accuracy.
• MD simulations and literature searches improved accuracy for density outliers.

Conclusions:

• Successfully developed accurate ML models for predicting organic compound density and refractive index (nD).
• These models facilitate informed decisions in organic process development and reduce wet lab work.
• The predictive capability is a significant step towards accurately forecasting solute solubility in various solvents.