Artificial Intelligence Predictions in Huge Chemical Spaces: Chiroptical Properties of [6]-helicene Family

Navigating the vast chemical space remains a major challenge in the rational design of materials with tailored properties. Here, we investigate how the properties of the [6]helicene family can be effectively modelled using a local, data-driven AI framework. By predicting each molecule from its closest structural neighbours, we accurately estimate diverse photophysical and (chir)optical properties. The coupling with genetic algorithms enables efficient inverse design and multi-objective optimization, yielding molecules unlikely to arise from intuition alone. The method uncovers [6]helicenes with enhanced electronic circular dichroism (ECD) features, tuned low-energy transitions, and exceptionally large g values, while revealing clear structure-property relationships that translate into practical design rules. Overall, this framework offers a general and efficient route for goal-directed molecular discovery across extensive chemical spaces.