摘要：AbstractDuring the past two decades, many computational tools were developed to aid novel biochemical pathway design. However, when longer pathways are to be predicted, putative reaction network with a large number of compounds will be generated after several steps of reactions. The combinatorial growth of the number of compounds makes expansion of the pathway search tree very slow and thus makes the identification of all possible pathways computationally intractable. Although methods based on chemical similarity have been developed to save computation time through reducing the number of nodes and branches in the reaction network according to their structural similarity to the target compound, the method can ignore important pathways containing compounds that are not relatively similar to the target compounds.We here report a new algorithm in our pathway prediction programAnnealPaththat improves the performance of pathway prediction in terms of the ability to predict pathways where the intermediates are not similar to the final target compound. We have used a global optimization algorithm, simulated annealing, to improve the efficiency of predicting longer non-linear pathways. We were able to show that our new algorithm is more computationally efficient, generates more possible pathways and identifies a higher number of shorter pathways.
