A Comparative Study of Predictive Coverage Based on Traditional Chinese Medicine Boinformatics Database and Analytics Tool Choices

As the demand for scientific validation of traditional medicine increases, network pharmacology research utilizing various bioinformatics databases has been actively conducted. However, clear guidelines on how differences in prediction outcomes may vary depending on the choice of database are lacking. This study aims to compare two major bioinformatics databases for herbal medicine (TCMSP, BATMAN-TCM) to analyze how the choice of database influences prediction results and to propose a more effective analytical approach. We compared the prediction results among three scenarios composed of different combinations from two TCM bioinformatics databases (TCMSP, BATMAN-TCM) and one analytics tool (DAVID) with an example herb, licorice. The gene ontology terms (GO-term) and Kyoto encyclopedia of genes and genomes (KEGG) pathway terms were compared and prediction ranges of two TCM bioinformatics databases for disease were compared against to disease list derived from in vivo study literatures of licorice written in the last 10 years. The three scenarios showed different trends in enrichment analysis with GO-term and KEGG pathway term. Scenario A showed a trend of cancer (apoptotic process, p=1.80E-32; response to hypoxia, p=8.90E-31, regulation of apoptotic process, p=2.30E-30, regulation of programmed cell death, p=6.40E-30, pathways in cancer, p=6.10E-36) whereas other two scenarios showed similar trend of neurotransmission (regulation of ion transport, p=1.00E-82, cell-cell signaling, p=9.16E-85, neuroactive ligand-receptor interaction, p=1.66E-37). The 58% and 52% of diseases derived from in vivo experiment study literatures were predicted by each TCM bioinformatics database. These results indicate that differences in target lists predicted from different databases lead to differences in enrichment analysis, which in turn leads to differences in disease prediction coverage. Thus, using a merged target list predicted by at least two TCM bioinformatics databases may provide more unbiased, complete, and wider range of results.