Accelerating\nthe Discovery of Abiotic Vesicles with\nAI-Guided Automated Experimentation

The first protocells are speculated to have arisen from\nthe self-assembly\nof simple abiotic carboxylic acids, alcohols, and other amphiphiles\ninto vesicles. To study the complex process of vesicle formation,\nwe combined laboratory automation with AI-guided experimentation to\naccelerate the discovery of specific compositions and underlying principles\ngoverning vesicle formation. Using a low-cost commercial liquid handling\nrobot, we automated experimental procedures, enabling high-throughput\ntesting of various reaction conditions for mixtures of seven (7) amphiphiles.\nMultitemplate multiscale template matching (MMTM) was used to automate\nconfocal microscopy image analysis, enabling us to quantify vesicle\nformation without tedious manual counting. The results were used to\ncreate a Gaussian process surrogate model, and then active learning\nwas used to iteratively direct the laboratory experiments to reduce\nmodel uncertainty. Mixtures containing primarily trimethyl decylammonium\nand decylsulfate in equal amounts formed vesicles at submillimolar\ncritical vesicle concentrations, and more than 20% glycerol monodecanoate\nprevented vesicles from forming even at high total amphiphile concentrations.