Integrated approaches to perform in silico drug discovery

Computer assisted, or in silico, drug discovery approaches play an important role in the search for small molecule hits and leads. These include structure- and ligand-based methods, as well as data mining and QSAR. They are used to analyze and predict ligand-receptor binding, as well as pharmacokinentic profiles of compounds with therapeutic potential. A diversity of offerings is publically/commercially available for performing these tasks. Each offering comprises select combinations of in silico methods. Efficient in silico drug discovery requires effective use of combinations of these tools. Unfortunately, no single vendor offering integrates all in silico capabilities. Typically, different vendors offer different "flavors" of the same method and specific "flavors" have associated strengths and weaknesses. Furthermore, significant inter-vendor format incompatibilities exist. Consequently, extensive scripting as well as manual intervention is required in order to overcome disparate data formats. In this article, we introduce the architecture and implementation of a highly efficient, and automated in silico drug discovery engine that integrates multi-vendor software. A single graphical user interface enables the user to 'Click & Configure' modeling tools and permits 'Mix & Matching' components from various vendors. It deploys a 'Divide & Conquer' strategy to marshal the resources of a multi-node compute cluster for compute-intensive tasks. This basic framework in performing in silico modeling activities (work-flow automation) envisions the integration of structure-based, ligand-based, and other modes of in silico drug discovery.