The human immunodeficiency virus (HIV), causing acquired immune deficiency syndrome (AIDS), is presently one of the most important pathogens affecting mankind. Worldwide, there were an estimated 1.8 million new infections and over one million AIDS deaths in 2016, according to the World Health Organization. HIV is more prevalent in the developing world, but also carries a large economic burden for wealthy countries by affecting young adults; over 15,000 people aged 13-29 were diagnosed in the US alone in 2016, according to the Center for Disease Control. While targeting viral enzymes (reverse transcriptase, protease, and integrase) resulted in effective drugs, the problem of resistance is negatively affecting long-term treatment. In particular, the class of non-nucleoside reverse transcriptase inhibitors is composed of a few molecules, and HIV can become resistant with a few key mutations.<br/>We now describe a workflow for the design of novel Non-Nucleoside Reverse Transcriptase Inhibitors (NNRTIs) against HIV using Assay Central, a collection of predictive Bayesian models deployed in a self-contained executable. Briefly, our approach defined by building an initial Bayesian model to make predictions for a virtual library, supporting prediction scores with docking and literature, testing compounds for activity against HIV, and feeding newly acquired data back into the original model.<br/>Data available through NIAID ChemDB (https://chemdb.niaid.nih.gov/) was leveraged to build an array of Bayesian models of inhibition of viral replication proteins as well as whole-cell inhibition specific to assay and host cell line. Various metrics are available to the user within the Assay Central executable, and test sets of recent literature were compiled to further validate model predictivity. Receiver Operator Characteristic scores range from 0.89 to 0.94 and Matthews Correlation Coefficients ranged from 0.64 to 0.75. These models, along with others using ChEMBL data, have driven the synthesis of several sets of analogs to date and the resulting in vitro data have been used for private model construction and identification of active molecules.

Facilitating intelligent HIV drug design with assay central

256th National Meeting (2018)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.