Computer-Aided Molecular and Product Design

Computer-aided molecular and product design is an interdisciplinary discipline that combines molecular property prediction, algorithmic search and process engineering to accelerate the development of functional chemicals and formulated products. At its core, this approach transforms the traditional trial-and-error workflow into a systematic exploration of molecular structures and formulation variables, guided by predictive models and optimisation algorithms. Key components include the use of physics-based or data-driven property estimators to evaluate thermodynamic, kinetic and safety characteristics of candidate molecules; mathematical programming techniques to navigate discrete and continuous decision spaces; and decision-support tools that integrate customer specifications, environmental criteria and economic objectives. By coupling molecular design with process or product formulation, this methodology enables the simultaneous selection of solvent structures, ingredient combinations and operating conditions that meet multiple performance criteria. Recent advances in machine learning, Bayesian optimisation and high-throughput experimentation have further expanded the scope of computer-aided design, facilitating the discovery of novel solvents for carbon capture, sustainable formulations for detergents and customised flavour and fragrance compounds. This synergy between molecular insight and computational optimisation holds broad implications for energy, pharmaceuticals, consumer products and green chemistry, offering a path to more efficient, cost-effective and environmentally responsible solutions.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Computer-Aided Molecular and Product Design publication trend

The graph below shows the total number of articles in computer-aided molecular and product design across all publications each year (not limited to Nature Index journals).

Technical terms

Computer-aided molecular and product design: A systematic approach that integrates molecular property prediction, optimisation algorithms and product formulation to accelerate the discovery and development of chemical products.

Mixed-integer nonlinear programming (MINLP): A class of optimisation problems featuring both continuous and discrete variables subject to nonlinear equations or inequalities.

Group contribution methods: Predictive models that estimate molecular properties by summing contributions associated with predefined structural fragments.

SAFT-γ Mie: A group contribution equation-of-state based on statistical associating fluid theory used to predict phase equilibria and thermodynamic properties.

Bayesian optimisation: A probabilistic optimisation technique that employs surrogate models to efficiently explore high-dimensional design spaces with limited evaluations.

Loss function: A mathematical expression that quantifies the deviation between predicted or observed performance and target specifications.

Multi-attribute decision-making: A set of methods for integrating multiple criteria into a single performance metric to guide decision support in complex design problems.

References

1. Outer approximation algorithm with physical domain reduction for computer‐aided molecular and separation process design. AIChE Journal (2016).
2. Computer-aided design of formulated products: A bridge design of experiments for ingredient selection. Computers & Chemical Engineering (2023).
3. Incorporating Machine Learning in Computer-Aided Molecular Design for Fragrance Molecules. Processes (2022).