Enhancing Activation Energy Predictions under Data Constraints Using Graph Neural Networks

Accurately predicting activation energies is crucial for understanding chemical reactions and modeling complex reaction systems. However, the high computational cost of quantum chemistry methods often limits the feasibility of large-scale studies, leading to a scarcity of high-quality activation energy data. In this work, we explore and compare three innovative approaches—transfer learning, delta learning, and feature engineering—to enhance the accuracy of activation energy predictions using graph neural networks, specifically focusing on methods that incorporate low-cost, low-level computational data. Using the Chemprop model, we systematically evaluated how these methods leverage data from semiempirical quantum mechanical (SQM) calculations to improve predictions. Our findings reveal that delta learning, which adjusts low-level SQM activation energies to match high-level CCSD(T)-F12a targets, offers the greatest accuracy improvement. However, this approach also requires more computational resources due to the complexity of transition state searches for each query of activation energy during model application. Transfer learning, which pretrains models on large datasets of low-level data, provided mixed results, particularly when there was a mismatch in the reaction distributions between the training and target datasets. Feature engineering, which involves adding computed molecular properties as input features, showed modest gains, particularly when incorporating thermodynamic properties. Our study highlights the trade-offs between accuracy and computational demand in selecting the best approach for enhancing activation energy predictions. These insights provide valuable guidelines for researchers aiming to apply machine learning in chemical reaction engineering, helping to balance accuracy with resource constraints.