COSMO-RS predictions of logP in the SAMPL7 blind challenge

We applied the COSMO-RS method to predict the partition coefficient logP between water and 1-octanol for 22 small drug like molecules within the framework of the SAMPL7 blind challenge. We carefully collected a set of thermodynamically meaningful microstates, including tautomeric forms of the neutral species, and calculated the logP using the current COSMOtherm implementation on the most accurate level. With this approach, COSMO-RS was ranked as the 6st most accurate method (Measured by the mean absolute error (MAE) of 0.57) over all 17 ranked submissions. We achieved a root mean square deviation (RMSD) of 0.78. The largest deviations from experimental values are exhibited by five SAMPL molecules (SM), which seem to be shifted in most SAMPL7 contributions. In context with previous SAMPL challenges, COSMO-RS demonstrates a wide range of applicability and one of the best in class reliability and accuracy among the physical methods.

Automated summary

Using the COSMO-RS method, we accurately predicted the partition coefficient logP for 22 small drug-like molecules between water and 1-octanol in the SAMPL7 blind challenge, ranking 6th overall among 17 submissions with a mean absolute error (MAE) of 0.57. Our method demonstrated wide applicability and high reliability compared to other physical methods in previous SAMPL challenges.