COSMO-RS based predictions for the SAMPL6 logP challenge.

Area of Science:

• Computational chemistry
• Physical chemistry
• Drug discovery

Background:

• Accurate prediction of physicochemical properties like octanol-water partition coefficients (logP) is crucial for drug development.
• The SAMPL6 (6th physical property blind challenge) provided a benchmark for evaluating prediction methods.

Purpose of the Study:

• To predict octanol-water partition coefficients (logP) for small drug-like molecules within the SAMPL6 challenge.
• To evaluate the performance of COSMO-RS theory implemented in COSMOtherm for logP prediction.

Main Methods:

• Utilized COSMOtherm, a computational tool based on COSMO-RS (Conductor-like Screening Model for Real Solvents) theory.
• Employed two prediction sets: COSMOtherm with FINE parametrization (hmz0n) and COSMOquick (3vqbi) with a fast σ-profile algorithm and machine learning correction.

Main Results:

• COSMOtherm (hmz0n) achieved the highest accuracy among all SAMPL6 submissions based on root mean squared error.
• COSMOquick (3vqbi) secured 3rd place out of 91 submissions, demonstrating strong predictive performance.

Conclusions:

• COSMO-RS derived molecular free energies in solution are of high quality for predicting logP.
• The employed computational methods show excellent potential for application in drug discovery and development.