Improving the Decisionmaking Process in Lead Compound Identification Using Bayesian Statistics

• Activity Number: 352
• Type: Contributed
• Date/Time: Tuesday, July 31, 2012 : 10:30 AM to 12:20 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Jo Wick*+
• Companies: University of Kansas Medical Center
• Keywords: high-throughput screening ; area under the curve

Abstract

The failure rate of drugs that enter clinical testing is reported to be as high as 90%. Factors related to failure include unanticipated metabolism and toxicity or insufficient efficacy. Since a significant portion of the costs associated with drug development arise from drugs that fail, developing a model for predicting the activity of new compounds in clinical study from preclinical data could have a significant impact. The National Institutes of Health Chemical Genomics Center has developed a quantitative high-throughput screening approach that extends high-throughput screening to derive the concentration-response profile of a compound during the primary screen. However, compounds are still being selected for further investigation based only on measures of potency-a concern since many active chemical entities do not show complete biological responses. We use Bayesian methods to derive estimates of efficacy and further to classify and rank curves. Matched animal data was used to investigate the predictive properties of characteristics including AUC, maximum effect and potency, and simulation studies show marked improvement in the positive predictive value of the process.