J. Fraaije, J.-W. Handgraaf
Culgi BV and Leiden University,
Keywords: coarse-grained modeling, honest broker, open data access, business support decision system
Summary:The success of molecular modeling in industry hinges on two aspects: (a) the awareness of what chemicals are present in the system of interest, (b) the possibility to predict a relevant property, with a research-cycle dictated by fast-moving product development in teams. Both items merit their own discussion but are also strongly linked. The lack of chemicals definition is especially disturbing: if one does not know what chemicals are present, then clearly any molecular-based model must fail from the onset. It may be surprising to the academic community, but getting the composition right is not trivial, to say the least. This is not only due to the mere complexity and hetero-dispersity of a typical industrial (consumer) product. It is even more important that in many cases locked-down business relations prevent the analysis of supplied materials, and in as many cases vendors are not permitted to know the chemicals specification of the end-user product. The common solution is to avoid modeling entirely, and resort to age-old trial-and-error (perhaps using high-throughput screening) while relying on vendor-supplied materials properties brochures for background information. In a modern way of doing things, all such information is typical stored in a business support decision platform. Rational design of products across the value chain is currently almost impossible but could be, in some future, a great asset to the modeling community and industrial product development in general. We will discuss several such cases and potential solutions derived from the pharmaceutical industries, such as the ‘Honest Broker’ or open data access. The second item, of finding a model rapidly and reliably is in a sense more scientific, and therefore more in the realm of algorithms in physics-based modeling and data-driven (machine-learning) approaches. But without the connection to business support decision platforms, scientific models can get lost fast, and factually could be considered irrelevant for product development. Here we discuss the future of the contrary: to make scientific models survive and thrive in such hard landscape. We will discuss one scientific approach in more detail, namely how to attach fast coarse-graining molecular modeling to in-house or external chemical databases, with examples taken from the home care and personal care industries. Culgi is a software and service company, sponsored by an international audience of organizations from a range of industrial sectors (personal and home care, oil, chemicals, materials). Its flagship product is the Culgi modeling platform, that unifies all relevant modeling methods (quantum, molecular, thermodynamic, informatics) in one software environment.