Condensed-Phase Molecular Representation to Link Structure and Thermodynamics in Molecular Dynamics

Molecular design requires systematic and broadly applicablemethodsto extract structure-property relationships. The focus of thisstudy is on learning thermodynamic properties from molecular-liquidsimulations. The methodology relies on an atomic representation originallydeveloped for electronic properties: the Spectrum of London and Axilrod-Teller-Mutorepresentation (SLATM). SLATM's expansion in one-, two-, andthree-body interactions makes it amenable to probing structural orderingin molecular liquids. We show that such representation encodes enoughcritical information to permit the learning of thermodynamic propertiesvia linear methods. We demonstrate our approach on the preferentialinsertion of small solute molecules toward cardiolipin membranes and monitor selectivityagainst a similar lipid. Our analysis reveals simple, interpretablerelationships between two- and three-body interactions and selectivity,identifies key interactions to build optimal prototypical solutes,and charts a two-dimensional projection that displays clearly separatedbasins. The methodology is generally applicable to a variety of thermodynamicproperties.

Automated summary

This study focuses on learning thermodynamic properties from molecular-liquid simulations using the Spectrum of London and Axilrod-Teller-Muto representation (SLATM).Through analysis of the preferential insertion of small solute molecules toward cardiolipin membranes and selectivity against a similar lipid, this methodology reveals interpretable relationships between two- and three-body interactions and selectivity, and provides a two-dimensional projection to display clearly separated basins.