Ranking effects of candidate drugs on biological process by integrating network analysis and Gene Ontology

There are high preclinical attrition rates in current drug discovery. The efficient assessment approach in the high throughout candidate drugs screening still needs great improvement. We propose two hypotheses. First, both drug action process and biological process can be converted to a common space of gene or gene product profiling. Second, the strength of drug action on biological process can be realized in the context of biological network. Based on the above hypotheses, we establish an algorithm termed Network-based Assessment for Drug Action (NADA) to assess the action strength of candidate drugs on certain biological processes. Then NADA is used to prioritize the effects of six compounds from traditional Chinese medicine on endothelial cell migration, a simple process defined by Gene Ontology, in the biological network specific for a given pathological process, angiogenesis. The computational results are subsequently tested by the experiment on the migration of Human Umbilical Vein Endothelial Cells in vitro. The experimental ranks for six compounds generally agree with the predicted output of NADA. NADA also outperforms the DAVID and meet/min methods in terms of the experimental orders, suggesting that the network topological features may have a key role in catching the mechanistic relationship between drug action and biological process. Hopefully, the progress of network biology approaches for deciphering complex diseases will further expedite the preclinical screening and accelerate the development of treatment modalities.