Analysis of High.ppt

Analysis of High-Throughput Screening Data C371 Fall 2004 Drug Discovery Process The key steps of drug discovery are: research - average 2 to 3 years pre-clinical testing - average 1 year clinical trial testing (involving human patients) - average 10 years regulatory approval - average 2 years Drug Discovery Process: Web Sites http://akosgmbh.de/Drug_discovery_process.htm /PPD_U7.htm INTRODUCTION HTS allows hundreds of thousands of compounds to be assayed very quickly HTS data characterized by: High volume High level of noise Diverse nature of the chemical classes involved Possible presence of multiple binding modes INTRODUCTION Select the most potent compounds to progress to the next stage Problems: Functional groups that interfere with the assay (e.g., fluoresce) Functional groups that react with biological systems Catch these with substructure and “drug-likeness” filters Techniques for Analysis of HTS Data Can’t use multiple linear regression or partial least squares as statistical tests Data sets are too large Data visualization Data reduction Data mining (if activity data is known) HTS Methodology Procedure: Measure activity at different concentrations for a subset of compounds Define IC50 (Inhibitory Concentration 50): the concentration of a material estimated to inhibit the biological endpoint of interest (e.g., cell growth, ATP levels) by 50% Solid pure sample that tests positively gets structure determined (hits-to-leads phase) DATA VISUALIZATION Need to display simultaneously large data sets with many thousands of molecules and their properties Typical software packages: Draw various kinds of graphs Color selected properties Calculate simple statistics HTS data sets may be divided into subsets to aid navigation SpotFire DecisionSite DecisionSite Examples / Features of Data Visualization Often combined with structure searching to find compounds with certain features Unsupervised methods – don’t use activity data Supervised methods – incorpora