Bridging the Gap between Differential Mobility, Log S, and Log P Using Machine Learning and SHAP Analysis

Abstract

Aqueous solubility, log S, and the water−octanol partition coefficient, log P, are physicochemical properties that are used to screen the viability of drug candidates and to estimate mass transport in the environment. In this work, differential mobility spectrometry (DMS) experiments performed in microsolvating environments are used to train machine learning (ML) frameworks that predict the log S and log P of various molecule classes. In lieu of a consistent source of experimentally measured log S and log P values, the OPERA package was used to evaluate the aqueous solubility and hydrophobicity of 333 analytes. With ion mobility/DMS data (e.g., CCS, dispersion curves) as input, we used ML regressors and ensemble stacking to derive relationships with a high degree of explainability, as assessed via SHapley Additive exPlanations (SHAP) analysis. The DMS-based regression models returned scores of R 2 = 0.67 and RMSE = 1.03 ± 0.10 for log S predictions and R 2 = 0.67 and RMSE = 1.20 ± 0.10 for log P after 5-fold random crossvalidation. SHAP analysis reveals that the regressors strongly weighted gas-phase clustering in log P correlations. The addition of structural descriptors (e.g., # of aromatic carbons) improved log S predictions to yield RMSE = 0.84 ± 0.07 and R 2 = 0.78. Similarly, log P predictions using the same data resulted in an RMSE of 0.83 ± 0.04 and R 2 = 0.84. The SHAP analysis of log P models highlights the need for additional experimental parameters describing hydrophobic interactions. These results were achieved with a smaller dataset (333 instances) and minimal structural correlation compared to purely structure-based models, underscoring the value of employing DMS data in predictive models.

Machine Learning-Assisted Exploration and Identification of Aqueous Dispersants in the Vast Diversity of Organic Chemicals

“…Using the LGBM model, feature importance of the input descriptors was performed using Shapley additive explanations (SHAP) to examine the relative contribution of each descriptor to the predicted result (Figure S7). , This examination revealed two basic findings: First, the experimental descriptors (CNT type, CNT quantity, and ratio of dispersant to CNT) as a whole contributed more significantly to the model predictions. Second, molecular descriptors which captured the overall features of the molecular species (e.g., BCUT2D, Chi, SMR_VSA, and AvgIpc) contributed more than those which described more specific aspects of the molecule, such as the number of atoms (Supporting Information).…”

Machine Learning Accelerating the Condition Screening of Ceftriaxone Sodium Anaerobic Co-Metabolic Degradation

“…Its core principle is to calculate the marginal contribution of features to the model output, then explain the “black box model” on both the global and local levels . SHAP builds an additive explanation model, all features are regarded as “contributors”, and the mechanisms of influence behind different influencing factors on the target value are explained by deconstructing the contribution value. − SHAP is a “model interpretation” package developed by Python. In this study, Python 3.11 was used with the XGBoost model.…”

Towards the Prediction of Drug Solubility in Binary Solvent Mixtures at Various Temperatures Using Machine Learning

“…As a data-driven approach, ML models can be trained on available solubility datasets to learn solute-solvent interactions for prediction of unmeasured solubility. To date, most of these ML models were developed to model the solute solubility in a speci c solvent [14][15][16][17][18][19][20][21][22][23][24], with a primary focus on aqueous solubility due to availability of extensive data. These solvent-speci c models generally deliver high accuracy because their design does not require them to account for variations between different solvents.…”