Personalized Warfarin Dosing by Genomics and Computational Intelligence

The objective of this project is to develop new techniques to incorporate genomic data into pharmacodynamic models to improve the dosing of chronically administered drugs. Specifically, the investigators look to improve warfarin therapy by decreasing the variability in the effect of this drug using information about the subjects genotype and computational intelligence. The investigators propose to achieve our objectives using a prospective, randomized, controlled clinical trial of a computer program that they will develop from both historical and prospective data. This computer program will be tested against a control group using standard warfarin dosing, and a group using standard dosing plus subject genotype. Warfarin dose and response data will be obtained from patients seen in the Louisville VA anticoagulation clinic. Following informed consent, subject genotype for cytochrome P450 allele 2C9 (2C9) and vitamin K epoxide reductase complex subunit 1 (VKORC1) will be determined. Other data routinely obtained to aid in warfarin dosing will also be recorded. Using this information, the investigators will develop many different models for warfarin dosing using incrementally more information. Each of these models will be tested using computer simulation until they have obtained the best model. This model will be used in a pilot study to test performance in real time. The results of the pilot study will then be used to power a final clinical trial of standard dosing, standard dosing and genetic information, computer dosing, and computer dosing plus genetic information.

The specific aims of this research are:

1. Determine the structure and the type of neural network model for predictions from historically obtained data. (Computer Model)
2. Prospectively develop an individualized neural network and nonlinear mixed effect modelling (NONMEM) model capable of predicting erythropoietin dosing for chronic in-center hemodialysis patients using adaptive techniques.
3. Develop computer programs based on neural computing that can be used in a clinical setting. (Computer Model)
4. Determine the utility of the computer programs prospectively in the clinical setting.