APPLICATIONS

The BioSciences Group of Fujitsu Computer Systems today unveiled a new technique for generating enhanced predictions for ADME/Tox research, consisting of a unique docking-based approach combined with off-the-shelf and purpose-built technologies, to develop viable and effective predictive models. The new approach was developed with a commercial collaborator focusing on drug-resistant infectious diseases and cancer. Key challenges facing in silico ADME/Tox researchers include limitations in mathematical model approaches such as QSAR, and also in database-oriented approaches. QSAR approaches are often molecule-centric, poorly representing underlying biology, as well as being sometimes difficult to validate or repeat. Additionally, proprietary algorithms may create compliance issues and limit application to customer-specific workflows. Database approaches that focus on available literature are constrained by the efficiency of the literature curation process, which can be labor intensive. Fujitsu has developed technology and methods to overcome these limitations, and create a robust, in silico predictive platform for ADME/Tox. Fujitsu utilized several datasets to validate predictions, including the National Cancer Institute diversity datasets for initial calibrations and error estimates, and a proprietary database available from a commercial collaborator, to validate the in silico platform. The ADME/Tox technique developed by the BioSciences Group consists of a computational workflow designed to build metabolite models, identify and model active sites, and then run the results through a "high throughput docking" algorithm to dock the database of metabolites predicted against all Cytochrome P450 (CYP) models. The end result of the process is a valuable metric for relative distances of docked configurations, to determine how much a compound and its derivative classes interact with differing CYP models. To reduce the immense computational resources required to perform this workflow, the Fujitsu technique leverages hardware and software solutions to streamline and accelerate the process. The technologies involved and their role in the workflow include: 1. ADMEWorks -- An ADAPT-based product used for testing initial model sets. 2. BioFrontier P450 -- A curated database for testing CYP interactions. 3. Neurosim L -- A Neural Network program to identify descriptors. 4. Support Vector Machine analyses -- Developed in R, used to classify and parse datasets. 5. Fujitsu in silico Screening platform (isS) -- High-throughput, high-precision computer simulation of protein-ligand docking and free energy of binding calculations using BioServer grid computing platform and proprietary simulation software. 6. Homology Modeling for Active Site Identification -- Utilizes sequence threading, multiple alignments and based on knowledgeable hashing protocols, for creating models of CYPs. 7. BioMedCAChe -- A suite of predictive modeling tools used to visualize small molecules and active sites, and for running ActiveSite, which uses a docking algorithm based on an extended version of PMF. 8. Ligand Interaction Distance (LID) Score -- Developed in-house, utilizing matrices of bumps and bonds and MASCs, and used to quickly score docked configurations across target types. "Developing and utilizing docking scores as descriptors provides biosciences researchers with a valuable and actionable metric for creating stronger predictive models," said Takahiro Tsunekawa, vice president of the BioSciences Group. "Fujitsu has created a powerful technique to leverage existing technologies and solutions to reduce computational and workflow burdens that have made this type of research prohibitive in the recent past." According to Dr. Michael J. McManus, vice president of Business Development for the BioSciences Group "Conventional ADME/Tox 'black box' approaches alone are not capable of telling researchers everything they need to know about the multifactorial processing involved in drug reaction in a living system. The Fujitsu view is that different problems are going to involve different levels of in silico modeling support and different amounts of wetlab validation." He added, "The value of the Fujitsu approach is that it provides experimental scientists with a model that allows them to get closer to the underlying biology, by providing a more 'complete' view of their research targets, as opposed to complex but limiting mathematical models. As this validation approach is built upon a foundation of Fujitsu existing in silico screening technologies and processes, researchers can benefit from a practical and affordable technique."