(630f) Molecular Product Design Under Performance Constraints

In recent decades, the growing awareness of environmental issues has prompted industries to seek sustainable alternatives for conventional chemical processes and products. Within the realm of surfactants solutions, their wide and extensive industrial applications underscore their growing significance of these systems in our daily lives, emphasizing the need for a systematic approach to tailor surfactant molecules to meet diverse applications demands while adhering to sustainability regulations. However, achieving this balance between optimal physical-chemical characteristics and environmental considerations in surfactant design is a complex task that requires a deep understanding of the structure-property relationships of the compounds involved (Alshehri et al., 2020; Samudra & Sahinidis, 2013). In this context, the integration of artificial intelligence and machine learning models has demonstrated remarkable capabilities to analyze and optimize complex systems to aid data-driven decision-making (Ajagekar & You, 2023; Moriwaki et al., 2018; Qin et al., 2021).

An optimization framework has been developed for the design of surfactants of industrial interest. This novel methodology is based on the individual optimization of the main parts comprising the surfactant along with multi-objective optimization of the whole surfactant properties. Predictive models are developed to predict the most pertinent properties of these products, including critical micelle concentration, synthetic accessibility score, Krafft point, surface tension, biodegradability and toxicity. This approach reduces the computational time required and the range of unrealistic molecular structures. Accuracy is improved by imposing constraints and specific desired attributes such as number of different molecular groups in head, minimum number of carbons in tail, etc. Furthermore, the framework enables the exploration of alternative structures diverging from the traditional configuration.

Two illustrative case studies are included to demonstrate the utility and effectiveness of the framework. Regarding structural optimisation, 102 heads and 50 tails are generated from the combination of 21 molecular groups instead of the 2²¹ possible structures that would be obtained without the added constraints. Databases of known surfactants that may be of interest can be added in this step. Continuing with multi-objetive property optimization, the first case study entails the maximization of biodegradability and minimization of both, cmc and synthetic accesibility score. The second case proposes the minimization of toxicity and synthetic accesibility score. Biodegradability and toxicity are greatly improved in the surfactants found as optimal solutions, such as different kinds of sugar-based surfactants, compared to the most typical ones in the industry. Additionally, the framework enables swift comparison between the properties of known surfactants of interest.

References

Ajagekar, A., & You, F. (2023). Molecular design with automated quantum computing-based deep learning and optimization. Npj Computational Materials, 9(1), 143.

Alshehri, A. S., Gani, R., & You, F. (2020). Deep learning and knowledge-based methods for computer-aided molecular design—toward a unified approach: State-of-the-art and future directions. Computers & Chemical Engineering, 141, 107005. https://doi.org/https://doi.org/10.1016/j.compchemeng.2020.107005

Moriwaki, H., Tian, Y.-S., Kawashita, N., & Takagi, T. (2018). Mordred: a molecular descriptor calculator. Journal of Cheminformatics, 10(1), 4. https://doi.org/10.1186/s13321-018-0258-y

Qin, S., Jin, T., Van Lehn, R. C., & Zavala, V. M. (2021). Predicting critical micelle concentrations for surfactants using graph convolutional neural networks. The Journal of Physical Chemistry B, 125(37), 10610–10620.

Samudra, A. P., & Sahinidis, N. V. (2013). Optimization‐based framework for computer‐aided molecular design. AIChE Journal, 59(10), 3686–3701.