Keywords: computational chemistry, homecare
Summary:Computational chemistry seeks to model the behavior of chemical compounds using data-driven or physics-based modeling. The industrial applications so far have focused on understanding behaviors (as in extending instrumental chemical analysis), and on product and process design. Recently, it has become apparent that computational chemistry could have a third area of application, namely that of patent analysis. Patent analysis is obviously right in the center of industrial research and development, affecting many areas of research in product and process development, sales, and marketing. Patent analysis is usually about landscaping: who is doing what, and what are the relations between various patents, all driven from a business development point of view. The extension with computational chemistry adds to such analysis physical chemistry insights, such as the automated parsing of ingredients, or the automated screening whether such compounds would support claims and to what extent. In the pharmaceutical industry, such content analysis exists already for quite some time, especially concerning reading molecular structures of active compounds. We adventured in the domain of personal care applications, where we believe such computational chemistry patent analysis did not exist before. The extension is challenging since in many cases personal care ingredients could be of natural or mixed origin, only to be identified by common names, and the application areas themselves are sometimes identifiable only through age-old descriptions of natural materials and processes that lack scientific justification. In our approach, we concentrate first on a few applications in specific areas, that all have to with the stability of emulsions and microemulsions. As ways of demonstration of the technology, we will take a few recent micro-emulsion patents in cosmetics and discuss the various stages of the analysis: the automatic parsing of ingredients, the construction of a stability model, the looping through claims, and conclusions drawn from the analysis. The modeling part of the analysis is a mixture of various methods: data-driven (Machine Learning or QSPR), and physics-based modeling (thermodynamics and coarse-grained simulations). In all cases the analysis is through a scripted (python) workflow, combining various elements from the Culgi scientific modeling library. We will also briefly discuss the potential extension of the technology to other domains, such as surface treatment. Culgi Bv is sponsored by an international network of industries, from personal and homecare, oil industries, and c chemical companies.