Anti-HIV activity prediction and enrichment identify novel human targets for Anti-HIV therapy

Abstract

Human immunodeficiency virus (HIV) relies heavily on the host proteins to facilitate its entry and replication. Currently, more than 4000 human proteins are recorded to be involved in the HIV-1 life cycle. Identifying appropriate anti-HIV targets from so many host proteins is crucial to anti-HIV drug development, but a challenging work. Here we combined anti-HIV activity prediction and enrichment analysis to identify novel human targets for anti-HIV therapy. We firstly developed an accurate prediction tool named Anti-HIV-Predictor (AUC>0.96) to predict the anti-HIV activity of given compounds. Using this tool, we predicted 10488 anti-HIV compounds from ChEMBLdb. Then, based on this result and relationships of targets and compounds, we inferred 73 anti-HIV targets that enriched with anti-HIV compounds. The functional annotation and network analysis revealed that they directly or indirectly interact with 20 HIV proteins through neuropeptide signaling, GPCR signaling, cell surface signaling pathway, and so on. Nearly half of these targets overlap with the NCBI HIV dataset. However, the percentage of known therapeutic targets in these targets is significantly higher than that in the NCBI HIV dataset. After a series of feature analysis, we identified 13 novel human targets with high potential as anti-HIV targets, the inhibitors of which have experimentally confirmed anti-HIV activity. It is noteworthy that the inhibitors of REN and CALCA have better anti-HIV activity than CCR5 inhibitors. Taken together, our findings provide novel human targets for the host-oriented anti-HIV drug development and should significantly advance current anti-HIV research.