Novel enhanced applications of QSPR models: Temperature dependence of aqueous solubility

• Published in: Journal of computational chemistry 2016-08, Vol.37 (22), p.2045-2051
• Main Authors: Klimenko, Kyrylo, Kuz'min, Victor, Ognichenko, Liudmila, Gorb, Leonid, Shukla, Manoj, Vinas, Natalia, Perkins, Edward, Polishchuk, Pavel, Artemenko, Anatoly, Leszczynski, Jerzy
• Format: Article
• Language: English
• Subjects: Approximation; Aqueous chemistry; aqueous solubility; Comparative analysis; feature net; QSPR; temperature-dependent; Thermodynamics
• ISSN: 0192-8651; 1096-987X
• DOI: 10.1002/jcc.24424

A model developed to predict aqueous solubility at different temperatures has been proposed based on quantitative structure–property relationships (QSPR) methodology. The prediction consists of two steps. The first one predicts the value of k parameter in the linear equation lgSw=kT+c, where Sw is the value of solubility and T is the value of temperature. The second step uses Random Forest technique to create high‐efficiency QSPR model. The performance of the model is assessed using cross‐validation and external test set prediction. Predictive capacity of developed model is compared with COSMO‐RS approximation, which has quantum chemical and thermodynamic foundations. The comparison shows slightly better prediction ability for the QSPR model presented in this publication. © 2016 Wiley Periodicals, Inc. Solubility in water is one of the key physico‐chemical properties which can vary due to temperature change. Since experimental determination of solubility can be difficult, expensive, and time‐consuming, QSPR modeling was used for organic compounds aqueous solubility prediction in temperature range 4–97°C. The feature net technique allows for the determination of the solubility parameter k from linear regression equation for better model performance. Models have acceptable predictive capability comparable to COSMO‐RS quantum chemical calculations.