Computational Predictions of Amphiphile Aggregation for Early Compartmentalization

Abstract

Extant biology uses a vast array of lipids to perform a range of tasks, and compartmentalization is critical for Life's existence by providing, a separation of chemical environments, enhanced local concentration of molecules, interfaces with reduced dimensionality, and individuality, leading to competition and evolution. We wished to explore and predict which kind of molecules are able to aggregate to form compartments that can host and/or encourage complex and perhaps even simple life-like chemistry that can be assayed easily in vitro. There may be a very large numberof such molecule types, and the use of high- resolution models is computationally prohibitive. We thus set out to develop an efficient way to predict aggregation and screen large in silico-generated compound libraries.

There are a range of methods available for producing or accessing libraries of molecules. Through the recent explosion of lipidomics, there are a number of tools developed for mass spectrometry that include large compound libraries (e.g. LipidBlast, LipidHome, etc.). These give access to biologically relevant lipids, but do not facilitate the identification of novel molecules. We have identified computationally cheap methods for the generation of exhaustive lipid libraries and the evaluation of their propensity to self-assemble into either micelles or vesicles. Depending on user-defined parameters such libraries can easily contain well past trillions of molecules. We used MolGen (http://www.molgen.de/) for exhaustive generation of sub-libraries of lipid tails and heads. MolGen allows for disallowed molecular motifs and ranges of molecule parameters to be defined for the output.

Once generated, solubility properties are assessed using QSPR models, and geometric properties computed. These are then combinatorially reacted using ChemAxon's Reactor software (https://chemaxon.com/) to give a final library. We finally evaluate them using chemoinformatics approaches to identify molecules that possess properties commensurate with an ability to form micelles, and more discriminatively, vesicles.