Simulations Plus, Inc. (Nasdaq: SLP), a leading provider of modeling and simulation software and services for pharmaceutical safety and efficacy, today announced that it has received initial experimental results from a collaborative research agreement it entered with a large pharmaceutical company in mid-2020 to evaluate the impact of the generative chemistry technology contained in the new Artificial Intelligence-driven Drug Design (AIDD) Module in ADMET Predictor®.
Working alongside the partner company’s R&D team, computational chemists at Simulations Plus evolved quantitative structure-activity relationship (QSAR) models for the specified drug target and defined a set of predicted endpoints against which to evaluate virtual lead molecules within the AIDD Module. The first 10 AIDD candidate molecules selected have now been synthesized and assayed by the partner. Besides being tested against the clinical target in whole cell assays, six of the compounds were evaluated for aqueous solubility and in rat and/or human microsomal clearance assays for in vitro metabolism. Two compounds were also assayed for binding to the human ether-a-gogo (hERG) gene product to assess their potential for cardiotoxicity.
Dr. Robert D. Clark, Senior Research Fellow at Simulations Plus and co-PI on the project, said: “To say we are very pleased with these results would be an understatement. Eight of the 10 designed compounds exhibited an IC50 below 1 µM, and the two most potent compounds had IC50 values below 100 nM. Perhaps just as importantly, we predicted the IC50 for biological activity quite accurately – our error was only threefold on average and within twofold for 60% of the candidates. All but one of the six solubilities measured were within fivefold of the ADMET Predictor estimated value, and the average predictive errors for rat and human microsomal clearance, also estimated using the ADMET Predictor default models, were 2.0- and 2.4-fold, respectively. More specifically, only one of the rat clearance predictions was off by more than threefold and none of the human clearance predictions were off by more than fivefold. This is a remarkable accomplishment for any drug design program.”
“The AIDD technology within ADMET Predictor enhances the dynamics of the medicinal and computational chemists’ interaction. Importantly, the program can adapt to synthetic chemistry requirements,” added Dr. Eric Jamois, Director of Business Development. “These results demonstrate the unique value of the ADMET Predictor machine learning models coupled with the powerful AIDD technology, which required only activity data to optimize the molecules across a wide spectrum of properties. We are thrilled with how successfully this collaboration has advanced the validation of the AIDD Module in ADMET Predictor to help find the proverbial ‘needle in a haystack,’ and we eagerly await the next set of results from the ongoing research program.”