We propose a fast and efficient method to compute molecular descriptors based on the Integral Equation Theory (IET) of Molecular Liquids. The new RISM-MOL-INF descriptors are obtained from solute-solvent correlation functions computed by the 1D Reference Interaction Site Model (RISM). The RISM-MOL-INF descriptors are conceptually different from those commonly used in chemical informatics, since they fingerprint solutes by the change they induce in solvent density distribution, rather than by the solute molecular structure alone. It is shown that key solvation thermodynamic and ADME properties can be predicted accurately using machine learning algorithms trained on RISM-MOL-INF descriptors only. Since they can be evaluated in a matter of minutes for most small organic and druglike molecules, the RISM-MOL-INF descriptors are suitable for medium to high-throughput in silico screening in applications such as pharmaceutical research and development. Due to the importance of solvation and desolvation effects in biological systems, we believe that the RISM-MOL-INF descriptors will find many applications in biophysical and biomedical property prediction.

Is there a role for the integral equation theory of molecular liquids in chemical informatics?

248th National Meeting (2014)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.