A Multiple Environmental Parameters and Molecular Fingerprints Contribution model for prediction of Gibbs free energy

Accurate prediction of thermodynamic parameters in biochemical reactions is essential for understanding and designing metabolic systems. Most existing methods for predicting the Gibbs free energy of biochemical reactions often neglect the environmental influences on Gibbs free energy such as pH, temperature and ionic strength, and lack efficient feature selection mechanisms, resulting in suboptimal predictive accuracy. In this paper, a Convolutional Neural Network Based Model with Multiple Environmental Parameters and Molecular Fingerprint Contribution (MEFC-CNN) is proposed to address these problems. Firstly, an encoding method that incorporates environmental factors is proposed to improve the ability to represent features. Secondly, a convolutional neural network with multiple parallel feature inputs is designed to efficiently select the key features, thereby improving the accuracy of Gibbs free energy prediction of biochemical reactions. Experimental results demonstrate that the MEFC-CNN model achieves superior predictive accuracy compared to existing methods.