- Aureus Sciences builds knowledge bases for predicting drug candidates' safety, focusing on areas like drug-drug interactions, safety pharmacology, and off-target effects.
- They have developed large structured databases of chemical and bioactivity information from literature and provide applications and services to analyze the data for customers in drug development.
- Their predictive models and databases have been shown to accurately predict drug interactions and off-target effects, helping customers optimize drug safety assessment.
9. Safety and Adverse Effects 31/03/11 Cliquez ici pour changer le texte Every year about 2 million patients in the United States are affected by a serious adverse drug reaction. This results in approximately 100,000 fatalities, making ADRs the fourth leading cause of death in the U.S. The consequence is a huge burden on national economies, which is estimated to be $136 billion annually in the U.S. Over the past 10 years, 19 broadly used marketed drugs were withdrawn after presenting unexpected severe side effects, with Rofecoxib and Cerivastatin.
10. DDI and Safety Problem? We need Predictive tools for safety assessment Examples of drugs withdrawn because of CYP-related DDIs See the US Food and Drug Administration (FDA) site ( http://www.fda.gov/medwatch/safety.htm ) for an updated list QTC Prolongation QTC Prolongation QTC Prolongation QTC Prolongation Toxicity Toxicity Toxicity Toxicity
11. Aureus Approach for Safety Prediction 31/03/11 Cliquez ici pour changer le texte
14. Quality Control: cross validation CHEMICAL MANUAL QUALITY CONTROL BIOLOGICAL MANUAL QUALITY CONTROL SCORE >5% NO NEW CROSS VALIDATION <5% YES Knowledge Database BATCH SAMPLE 50%
15. Quality Control : Statistical Approach Sql request: Definition of the set of data responding to specific criteria (parameter, protocol, substance, target…) Robust Grubbs Test ISO 5725-2 OUTLIERS Decreased number of false positives Multiplication factor to median
19. 31/03/11 Cliquez ici pour changer le texte Usage of Knowledge Base for Safety Prediction
20. Prediction Of DDIs : Approach DDI? AurSCOPE ADME/DDI Laboratory Marketed and Preclinical drugs In vitro Data Metabolism Inhibition PK Data Animal Human In vitro Data Metabolism Inhibition PK Data Human My drug IND Phase 1 Phase 2 Phase 3 NDA Phase 4 Discovery DDI Predict Profile of DDI (~100) In (2-3 minutes)
21. Aureus Pharma Confidential All CYPs Data Needed for DDI predictions Competitive inhibition Metabolism by different pathways (fm(E)) Mechanism-based inactivation (MBI) Inhibitor concentration [I] at the enzyme active site : - [I]in,max, [I]in,avg, Cmax, Cavg - fraction unbound in plasma (fu b ) - Non equilibrium conditions (Kp,Kb, C/M) CYP3A Inhibition of gut « first pass » metabolism (F’g/Fg) Competitive inhibition Metabolism by different pathways Mechanism-based inactivation (MBI)
22. Biological diversity for CYP3A and CYP2D6 inhibitors 31/03/11 Needs of structured data for retrieving in vitro and in vivo parameter values
23. 31/03/11 f m : Contribution of a particular enzyme to the overall clearance AUC ratio Sensitivity fm(E) Variability impact on DDI Predictions
24. Prediction Accuracy 31/03/11 MSM : DDI Predict 2009 Edition MSM DDI predict 2009 Percent Correct (>2 fold DDI <2 fold no DDI) 81 Percent within two fold 76 Percent under predict 4 Percent over predicted 20 MFE 1.69 RMSE 5
25. Safety Pharmacology 31/03/11 Cliquez ici pour changer le texte Off Targets are unwanted pharmacology activity of drug candidate on various targets related to adverse effects . The hERG-related K+ channel, the 5-HT2B receptor, or the PXR nuclear hormone receptor are typical examples which fall into this category. Extensive in Vitro Safety Pharmacology Profiling can be quite expensive (in the order of about $3000-$10,000 for a panel of 50-100 targets. To offset this there is increasing interest in in silico methods, which do not require any underlying functional knowledge, to predict possible ADRs
29. Prediction Accuracy using AurPASS Training Set: 47 938 molecules 517 Activity Types Mean Accuracy of Prediction: 98% Test Set: 2244 molecules 113 Activity types Mean Accuracy of Prediction: 90%