Abstract: Consumers increasingly demand more eco-friendly cosmetic products made from sustainably-sourced ingredients. However, widespread adoption of new sustainable formulations requires that they perform well relative to their synthetic counterparts. Artificial intelligence (AI) and machine learning (ML) have emerged as a disruptive technology in fields across industry from finance to medicine. Any application area where large amounts of data are collected or generated stand to benefit from the insights provided by ML by efficiently driving research direction and informing decisions. The role of ML in surfactant-based formulation development in the cosmetics industry is an active area of research and its potential has yet to be fully realized.
In this talk, we advocate for the use of active learning combined with molecular simulation in order to efficiently identify the most promising formulations for experimental validation. In addition to the variability in molecular structure of individual naturally-derived ingredients, mixtures of these ingredients into complex formulations opens a wide design space that is impossible to search exhaustively with Edisonian experimental approaches. Molecular simulation enables one to study and compute physical properties of virtually any conceivable formulation. Active learning directs the conditions of the in silico experiments in order to iteratively optimize robust ML models with predictive ability covering all of design space. In this talk, we will demonstrate the strength of this approach using the eco-friendly example of formulations made with rhamnolipid biosurfactants.
