Advancing Aqueous Solubility Prediction: A Machine Learning Approach for Organic Compounds Using a Curated Data Set.

Area of Science:

• Computational chemistry
• Chemical informatics
• Drug discovery

Background:

• Aqueous solubility is a critical physicochemical property influencing compound applications in drug development and materials science.
• Accurate prediction of aqueous solubility is essential for efficient chemical research and development.

Purpose of the Study:

• To develop and validate a novel machine learning model for predicting the aqueous solubility of organic compounds.
• To enhance the accuracy and generalization capabilities of solubility prediction models.

Main Methods:

• A curated dataset was merged from four sources, encompassing diverse organic compounds.
• Multiple machine learning and deep learning models were employed, integrating chemical descriptors, fingerprints, and functional groups.
• The finalized model was rigorously tested on the Huuskonen dataset (1282 compounds).

Main Results:

• The developed model achieved a high predictive performance with an R-squared (R²) value of 0.92.
• The model demonstrated a low Mean Absolute Error (MAE) of 0.40.
• The model outperformed existing prediction methods on a highly diverse dataset.

Conclusions:

• The novel machine learning model provides a highly accurate and reliable method for predicting aqueous solubility.
• This approach offers significant advantages over existing methods, particularly for diverse chemical compound sets.
• The model has strong potential for application in drug discovery and materials science.