EnTdecker - A Machine Learning-Based Platform for Guiding Substrate Discovery in Energy Transfer Catalysis.

Area of Science:

• Computational Chemistry
• Catalysis
• Machine Learning

Background:

• Identifying novel substrates for energy transfer (EnT) catalysis is hindered by the vast chemical space and limitations in current experimental and computational methods.
• Existing strategies for substrate discovery in EnT catalysis are often time-consuming and costly, limiting rapid advancement in the field.

Purpose of the Study:

• To develop an accelerated approach for discovering new substrates in EnT catalysis.
• To introduce the EnTdecker platform for large-scale virtual screening of potential EnT substrates.
• To leverage machine learning (ML) for predicting excited state properties crucial for EnT catalysis.

Main Methods:

• Creation of a comprehensive dataset of over 34,000 molecules relevant to EnT catalysis.
• Training of predictive machine learning models using this dataset to estimate excited state properties.
• Validation of the platform's efficacy through rediscovery of known substrates and experimental luminescence-based screening.

Main Results:

• The EnTdecker platform successfully identifies potential substrates for EnT catalysis through ML-based virtual screening.
• The platform demonstrated practical utility by rediscovering known successful substrates from existing literature.
• Experimental validation confirmed the predictive power of the ML models, showcasing reduced computational effort for property prediction.

Conclusions:

• The EnTdecker platform significantly enhances the efficiency of substrate selection for EnT catalysis.
• This tool is expected to increase the success rate of experimental investigations in EnT catalysis.
• EnTdecker provides a publicly accessible web application (entdecker.uni-muenster.de) to facilitate broader research in the field.