Cgip-Deepreac+ Model for Organic Chemical Reactions Based on Deep Active Learning

In synthetic organic chemistry, finding and optimising reaction conditions is a complex and time-consuming process. With the development of deep and active learning techniques, we have tools to predict and guide this complex process. The CGIP-DeepReac+ model introduced in this paper is a framework that combines deep learning and active learning strategies, designed to solve the problem of reaction prediction and optimisation in synthetic organic chemistry. CGIP-DeepReac+ is a synthetic organic chemistry prediction model based on deep active learning, which employs a graph neural network to deal with complex molecular structures, and combines with active learning strategies to Efficient selection of training data significantly improves the prediction performance. The model performs well in a number of organic chemical reactions, accurately predicting reaction results and optimising reaction conditions, providing a powerful tool for synthetic chemists. With data accumulation and model optimisation, CGIP-DeepReac+ is expected to play an important role in synthetic chemistry. The contribution of this paper is to demonstrate the potential of CGIP in enhancing the performance of the DeepReac+ model and to provide a new perspective on chemical reaction prediction. Future work will focus on further optimising the model structure, extending the dataset and exploring the application of CGIP in other chemical image analysis tasks.