Open Competition 1 - Information Technology

Pharmacogenomic Prediction Drug Therapy

Sponsor: Prediction Sciences, LLC.

• Project Performance Period: 12/1/2002 - 11/30/2005
• Total project (est.): $3,156,280.00
• Requested ATP funds: $2,000,000.00

Incorrect dosing of prescription drugs is the fifth-leading killer of Americans, resulting in about 7,000 deaths annually. An additional 2.2 million people suffer drug-related injuries or harmful side effects. At the same time, it is enormously expensive to develop a new drug, requiring about $1 billion from compound identification to marketplace release. This leads to the development of "all-or-nothing" blockbuster drugs. On the clinical side, a doctor spends, on average, an hour a day selecting and dosing drugs. This time could be better spent on patient care and, additionally, contributes to higher medical bills. One way to combat these problems while also controlling soaring medical costs would be to personalize drug therapy based on knowledge of how genetic variations and patient characteristics affect drug absorption, metabolism, and function. Prediction Sciences plans a three-year project to develop and test genetic and proteomic assays and information processing techniques for advanced decision-support tools that will enable physicians to predict drug reactions and tailor dosage ranges on a patient-specific basis. Information about biological pathways and a patient's genetic, medical, and environmental variables will be combined in a single predictive model, a challenging task because the relationship of these pathways and variables to specific drug reactions is not yet well understood. Medical data and blood samples for genetic analysis (without patient identifiers) already are being collected for use in developing and testing models. The researchers will identify small differences in particular genes, such as those related to metabolism and disease, and assemble them into clinically relevant combinations, which will be processed along with patient characteristic data to predict drug response and dosage. The proposed technology will be powerful because of its capability to analyze and associate large numbers of variables using advanced learning algorithms, and dynamically adaptive because it will become increasingly accurate as more patient data are input over time. In addition to the basic software, genetic assays will be designed for applications in psychiatry, cardiovascular disease, sepsis treatment, and immune-mediated disorders. Patient samples will be obtained from several University of California medical schools; Duke University, the University of Tennessee; Genomics Collaborative, Inc.; and EGreen, Inc. The ATP funding is needed because the research is too risky to attract venture capital investments. If successfully developed and commercialized, the predictive models will reduce health care costs, enhance the quality of care, and reduce drug development costs through better selection of patients for clinical trials. In addition, the models will increase physician productivity by reducing time wasted adjusting patient medications, and significantly improve life for patients suffering from diseases that today can require over a year of trial and error to find the optimum medication dosage.

For project information:

Cornelius Diamond, (619) 255-8730
cdiamond@predict.net

ATP Project Manager

Christopher Currens, (301) 975-8503
christopher.currens@nist.gov