Integrating herb effect similarity for network-based herb target prediction

Network pharmacology has become the new approach for drug mechanism research and novel drug design. Drug target prediction based on computational approach became one of the primary approaches. However, due to the diversity and complexity of herbal chemical structures, the performance of herb target prediction based on chemical structure similarity is limited by the quality and the data availability of herb chemical ingredients and their structural properties. To gain insights into the molecular mechanism of herbs by using clinical herb efficacies, we develop a computational approach to predict the potential targets of herbs by integrating the herb effect properties. We found that herbs with high effect similarities have high degree of shared targets. Meanwhile, an algorithm integrating propagation on protein-protein interaction network and effect-based herb similarity was proposed and obtained better accuracy than the chemical structure similarity. Furthermore, we manually evaluated some novel predictions like the target SP1 for herb turmeric, which is not recorded in the benchmark set, but has been confirmed by recent published paper.