Multi-lab intrinsic solubility measurement reproducibility in CheqSol and shake-flask methods

Area of Science:

• * Physical Chemistry
• * Pharmaceutical Sciences
• * Computational Chemistry

Background:

• * Accurate measurement of drug intrinsic solubility (log S₀) is crucial for drug discovery and development.
• * Existing datasets rely on various experimental methods, including the saturation shake-flask (SSF) method, which is considered a gold standard.
• * The CheqSol method offers an alternative approach for determining solubility values.

Purpose of the Study:

• * To compare the intrinsic solubility values obtained by the CheqSol method with those obtained by the SSF method.
• * To evaluate the performance and reliability of the CheqSol method in terms of accuracy and precision.
• * To assess the potential utility of CheqSol data for augmenting drug solubility prediction models.

Main Methods:

• * Comparison of 233 CheqSol log S₀ values from the Wiki-pS₀ database for 145 druglike molecules.
• * Analysis of 838 SSF log S₀ values for 124 molecules also present in the CheqSol set.
• * Statistical analysis including correlation (r²), root-mean-square error (RMSE), and bias assessment.

Main Results:

• * High correlation between CheqSol and SSF methods (r² = 0.96, RMSE = 0.34 log unit) with minimal bias (-0.07 log unit).
• * CheqSol demonstrated superior interlaboratory standard deviation (SD_i^CS = 0.15) compared to SSF (SD_i^SSF = 0.24).
• * Intralaboratory errors for CheqSol require a factor of 3 to align with expected interlaboratory errors, indicating potential systematic errors.

Conclusions:

• * The CheqSol method provides reliable intrinsic solubility data highly correlated with the SSF method.
• * CheqSol exhibits better interlaboratory precision than SSF, making it a valuable tool for solubility assessment.
• * Multi-laboratory averaged CheqSol solubility data can enhance training sets for predictive models in drug discovery.