Introduction
The performance of R&I (Research and Innovation) within industrial groups is always a hot topic, but especially in times of cost reduction. One of the ways to improve this performance is to make better use of available data, in the hope of reducing the need for testing. This hope has already been guiding R&I strategies in sectors such as oil & gas and pharmaceuticals for at least a decade. It is now arriving in force in the beauty sector.
New chemical R&D methods arrive in the beauty sector
Limitations of empirical approaches
Chemistry is at the origin of a large number of revolutionary drugs, but in recent decades the productivity and profitability of R&D based on empirical methods have been declining. In the case of the pharmaceutical industry, the cause of this phenomenon lies to a large extent in the fact that the failure of a molecule under development is becoming more and more expensive. However, all industrial sectors involved in the development of new molecules are confronted with the need to reduce development costs and lead times.
This reality leads sectors such as the oil & gas and pharmaceutical industries to actively seek to advance the methods of development of new molecules, by all means as long as they minimize the need for real tests. The goal is therefore to get as close as possible to a development environment that is as virtual as possible: an environment based on the manipulation of already available data. Why indeed launch an expensive experiment when you can simulate?
Challenges of data-driven methods
But do these data-driven approaches work? It depends on the complexity of the molecules and the systems with which they interact. Thus, the transition from in vitro to in vivo tests is a factor of complexity that is absent in the artificial environments characteristic of oil & gas.
Specificities of the beauty sector
In the case of the beauty sector, a new factor appears the complexity of formulations. Beauty products typically combine substances, particularly in the form of emulsions (the most complete and complex case). This reality makes the behavior of these formulations difficult to predict and can lead to uncertainty about the reliability of the products over time. A manufacturer cannot afford to market a product that does not keep its promises over time.
To increase the performance of R&I in the beauty sector, two data-driven approaches seem particularly interesting: retro-engineering on the one hand and predictive stability typology on the other.
Le rétro-engineering data-driven
This method consists in relying on data to correlate performances and properties with the structure of the molecules, and thus provide aid to the formulation (which is then no longer solely empirical). This method is located upstream in the design phase of new molecules and allows to limit the use of empirical tests. It, therefore, complements the work of the formulator. Companies in the beauty sector are showing great interest in this method, which has already been successful in the pharmaceutical industry, thanks in particular to advances in systems biology.
Data-driven prediction of formulation stability
Limitations of existing tests
Unlike a sector such as the food industry, the beauty sector is not subject to a regulatory obligation in terms of testing to guarantee the stability of formulas. Frédéric Leroy, Fellow Presans and former Director of R&D at L’Oréal Advanced Research draws attention to the fact that the main test in force only concerns the physical aspect (texture) of stability (the test consists of keeping the formulation at 45° for three months). This type of test has no predictive value. In particular, it does not give any information on the chemical aspect of stability: do the compounds degrade (disappearance of desirable compounds)? The same applies to the microbiological safety aspect (appearance of undesirable compounds). Consequently, the expiry dates indicated on the packaging of cosmetic products tend to be largely conventional.
In the case of formulas combining various substances, a better knowledge of their behavior would also contribute to reducing the risks and delays associated with the development of new molecules (faster elimination of false leads).
Using artificial intelligence to build a predictive typology
To build up this knowledge, it is necessary to compile a mass of data that currently exists only in a widely dispersed state. Such work would make it possible to constitute a predictive typology, combining tests and modeling, and making it possible to know in advance the effects of various combinations of substances. The ability to predict the evolution of a formula as a function of standard parameters (time, temperature, pressure) would increase massively, on the three levels of physical, chemical, and microbiological safety. However, the complexity and non-linearity of the formulas make it difficult to establish global laws of behavior.
Conclusion
The shift to data-driven R&I in the beauty industry is based on the adoption and adaptation of promising approaches that have proven successful in other industries. The question remains as to how such a knowledge base is to be built: will everyone work individually to strengthen their competitive position through the constitution of proprietary knowledge? Or is it possible that synergies may emerge between different players, or even between players belonging to different sectors, in a Synergy Factory logic?
