Enhanced ADMET Predictions Using High Throughput Docking Scores as Descriptors
BioSciences Group, Fujitsu
In silico ADMET modeling has taken an increasingly significant place within the drug development pipeline. The key limitation to developing high quality models to-date has been limitations in the descriptors available for inclusion into predictive tools coupled to throughput bottlenecks associated with key validation steps in the process.
For example topological indices, while useful for general 2D parsing of compound classes, tend to underestimate true 3D effects acting upon compounds in vivo, while electronic approaches may underestimate the true structural diversity present within varying compound classes. In addition, the complex signaling network within a living cell can make accurate predictions exceptionally challenging. For example, off target effects of a compound and/or its metabolites can be exceptionally challenging to predict, especially across varying protein subfamilies like the CYPs.
Scientists at Fujitsu have developed an integrated high throughput platform which ameliorates current limitations to ADMET modeling accuracy and throughput. The approach includes a combination of classic QSAR-based workflows identifying topological, quantum mechanic and semi-empirical quantum mechanic descriptors utilizing Scigress Explorer (BioMedCAChe), high throughput (highly parallel) protein docking (PMF-based FastDock) utilizing the BioServer, highly automated in situ hybridization utilizing Large Scale in situ Hybridization (LisH) and automated high throughput cellular injection utilizing MICAN.
Significant improvements in throughput can be achieved with such technologies. For example, FastDock running on BioServer can increase docking-based screening throughput up to 50 times over conventional computational methods, while LisH can increase target validation and safety screening up to 30 times over conventional manual methods. Similarly, MICAN can increase lead validation up to 100 times over manual injection methods.
A ) With BioServer, scientists can use High Throughput Docking to screen very large compound libraries against modeled active sites with unmatched throughput, enabling rapid detection of the most promising lead candidates. B) LisH enables high throughput analyses of target classes, providing in vivo validation and context to inform the selection of targets. C) Desktop modeling of a selected target is accelerated with Scigress Explorer (BioMedCAChe), enabling Structure Based Drug Design workflows. D) MICAN brings automation to in vivo validation studies, allowing rapid testing of the most promising lead compounds in a variety of cell types.
Ian G. Welsford, Ph.D. Fujitsu Computer Systems, BioSciences Group.