Biotechnology HPC Software Applications Institute (BHSAI)

BHSAI Develops a Website for Predicting Pharmacological & Toxicological Properties of Molecules Not Studied Before

June 29, 2018   |  Download PDF

Military medicine is facing ever increasing and greater challenges in maintaining Force Health Protection by diagnosing and treating Warfighters exposed to diseases and hazardous chemicals. The threat of a chemical or toxicant being released among civilians during a terrorist attack creates a strong incentive to develop effective countermeasures and tools to quickly assess the pharmacological and toxicological properties of compounds not studied before.

The U.S. Army Medical Research and Materiel Command is involved in developing drugs and countermeasures in multiple areas, including military-relevant infectious diseases and chemical biological defense. Drug discovery is a risky and resource-intensive endeavor with high attrition rates. For a drug to be effective, a potent molecule must reach its target in sufficiently high concentration and remain in its bioactive form long enough to have an impact. For almost three decades, a major focus in drug discovery has been to reduce the attrition rate. The development of assays and mathematical models to assess the adsorption, distribution, metabolism, excretion, and toxicity (ADMET) properties of a compound has greatly reduced the attrition rate. Yet toxicity remains a hurdle, and is responsible for 40% of failures among new drugs.

In an effort to reduce the attrition rate and identify ADMET issues at an early stage, TATRC’s Biotechnology High Performance Computing Software Application Institute (BHSAI) is working together with the Defense Threat Reduction Agency to develop mathematical models that predict ADMET properties. A team of scientists at the BHSAI, led by Dr. Anders Wallqvist, has developed the variable Nearest Neighbor (vNN) method, which can rapidly assess the molecular properties of a compound with high accuracy. The vNN method relies on the basic idea that structurally similar molecules have similar physical properties and biological activity. “The fundamental idea has been known for more than a century, when Overton and Meyer discovered a correlation between the lipid solubility for a series of compounds and their anesthetic potency,” said Dr. Patric Schyman, a scientist at the BHSAI. “But the ramifications of the vNN method go beyond just predicting molecular properties; it does so with confidence by only making predictions for molecules that are similar to the reference molecules included in the model – this is the model’s applicability domain.”

The ADMET predictions result page graphicThe ADMET predictions result page. The 15 ADMET predictions for each query molecule are presented on a separate row. High confidence predictions are shown in solid colors and low confidence predictions are shown in striped. The website can be publically accessed at  

Dr. Schyman developed and implemented the 15 ADMET in silico models, which are available on the vNN-ADMET website, and Mr. Valmik Desai designed the web-based user interface, which is capable of analyzing thousands of molecules. Dr. Schyman commented that “the vNN-ADMET web-tool will assist drug developers in rapidly evaluating pharmacological and toxicological properties in silico, before prioritizing and synthesizing compounds that have the highest chance of succeeding in the late stages of development.” A description of the website and the method was recently published in Frontiers in Pharmacology.

Schyman P, Liu R, Desai V, Wallqvist A. vNN web server for ADMET predictions. Frontiers in Pharmacology. 2017; 8: 889.

This article was published in the June 2018 issue of the TATRC Times.