GEMS - Guided Evolutionary Molecule Design for Sustainable Chemicals
Authors
Coelina Robinson (ETH Zurich), Franziska Weissbach (ETH Zurich), Kjell Jorner (ETH Zurich), Mennatallah El-Assady (ETH Zurich), Christina Humer (ETH Zurich)
Presentation
- Session
- Big Data, Bigger Physics
- Time
- Friday, Nov 13, 09:03 – 09:12 (US/Eastern) · session 08:00 – 09:30
- Location
- Hall America south
Keywords
Human-in-the-loop optimization, molecular design, genetic algorithms, green chemistry, safe-and-sustainable-by-design, interactive visualization
Abstract
Designing safe and sustainable chemicals is critical to combat chemical pollution in our environment. Computational and AI-assisted methods have been developed to aid de novo molecule design. However, data on the environmental impacts of chemical compounds are sparse, resulting in low-fidelity machine learning (ML) oracles and unreliable candidate proposals. Furthermore, many automated molecular design approaches rely on numerical scoring functions that cannot fully capture the nuanced chemical intuition of expert scientists required for real-world molecular design. Instead, we present GEMS—an interactive visual analytics tool for human-in-the-loop molecular optimization that lets domain experts directly collaborate with an evolutionary genetic algorithm. Users continuously guide the search using domain knowledge through high-level, parametric modification of the scoring function alongside direct, granular control over molecule populations. GEMS requires no programming or expertise in ML or evolutionary optimization. A usage scenario demonstrates its application in designing sustainable antioxidant alternatives, and interviews with domain scientists provide feedback on its usefulness.
For Practitioners
This paper is relevant to computational chemists, environmental chemists, molecular designers, and researchers developing safe-and-sustainable-by-design (SSbD) chemicals. It is also of interest to practitioners working with AI-assisted molecular design, evolutionary optimization, and interactive visual analytics. Practitioners can apply the concepts presented in this work to redesign molecular optimization workflows around continuous human involvement rather than fully automated optimization. By integrating expert knowledge throughout the search process, they can iteratively refine scoring objectives, identify and correct failures such as reward hacking, and steer candidate generation toward chemically meaningful solutions. The presented workflow also demonstrates how interactive visual analytics can reduce dependence on developers, making AI-assisted molecular design more adaptable to new application domains and evolving research goals.