MBPathNCP: A Metabolic Pathway Prediction Model for Chemicals and Enzymes Based on Network Consistency Projection

Background: Metabolic pathway is an important biological pathway in living organisms as it produces necessary energy to maintain vital movement. Although main part of metabolic pathway has been uncovered by the great efforts in recent years, its completeness is still a problem. The undetected chemical reactions in metabolic pathway have become a hinder for better understanding on its mechanism. Prediction of metabolic pathways that a chemical or enzyme can participate in is the first step to remove this hinder. Objective: This study aimed to design an effective computational method to predict the metabolic pathways of chemicals and enzymes. Methods: A new computational model was proposed to predict the metabolic pathways of chemicals and enzymes, which was called MBPathNCP. The kernels for chemicals/enzymes and pathways were constructed using the interactions of chemicals and proteins, and the validated associations between chemicals/enzymes and pathways. The network consistency projection was applied to the kernels and association adjacency matrix to yield the association score for each pair of chemicals/ enzymes and pathways. Results: Cross-validation results on this model shown its good performance. The further tests indicated the reasonability of the entire architecture and its superiority when the negative samples were much than positive samples. Conclusion: The proposed model MBPathNCP was efficient to predict the metabolic pathways of chemicals and enzymes and can be a latent useful tool to investigate metabolic pathway system.