Multiple linear regression models for predicting the n‑octanol/water partition coefficients in the SAMPL7 blind...

Area of Science:

• Computational chemistry
• Quantitative structure-property relationships (QSPR)

Background:

• The n-octanol/water partition coefficient (log P) is a critical physicochemical property for drug discovery and environmental fate assessment.
• Accurate prediction of log P is essential for screening and prioritizing chemical compounds.

Purpose of the Study:

• To develop and evaluate a multiple linear regression (MLR) model (MLR-3) for predicting the log P of N-sulfonamides.
• To assess the performance of the MLR-3 model in the context of the SAMPL7 blind challenge.

Main Methods:

• A multiple linear regression (MLR-3) model was developed using a training set of 82 diverse organic molecules.
• The model was applied to predict the log P of 22 N-sulfonamides in the SAMPL7 blind challenge.
• Model performance was evaluated using root-mean-square error (RMSE) and mean absolute error (MAE).

Main Results:

• The MLR-3 model, submitted as "TFE-MLR", achieved a RMSE of 0.58 and MAE of 0.41 log P units.
• The model demonstrated the highest accuracy among all empirical methods and all submissions in the SAMPL7 challenge.
• 75% of ranked empirical submissions achieved RMSE < 1 log P unit, highlighting the effectiveness of these methods.

Conclusions:

• The MLR-3 approach is highly appropriate for computing the n-octanol/water partition coefficient of sulfonamide-containing compounds.
• Empirical methodologies, including MLR, are suitable for fast and accurate predictions of physicochemical properties like partition coefficients for bioorganic compounds.
• The study validates the utility of QSPR models in predicting key properties for chemical and pharmaceutical research.