|
Activity Number:
|
356
|
|
Type:
|
Invited
|
|
Date/Time:
|
Wednesday, August 1, 2007 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Physical and Engineering Sciences
|
| Abstract - #307872 |
|
Title:
|
Multiple Mechanisms from a High-Throughput Screen
|
|
Author(s):
|
Douglas M. Hawkins*+ and Lexin Li
|
|
Companies:
|
The University of Minnesota and North Carolina State University
|
|
Address:
|
313 Ford Hall, Minneapolis, MN, 55455,
|
|
Keywords:
|
Mixtures ; Inverse regression ; Dimension reduction
|
|
Abstract:
|
Compound libraries in high-throughput screening often comprise groups of compounds operating by different mechanisms, a feature that bedevils methods that fit a single model to the entire collection. Recognizing the different mechanisms holds the potential for substantial improvement in the screening process. Recursive partitioning was shown some 10 years ago to have the potential to recognize this situation, but is conceptually not the best tool. Promising approaches include linear mixture models. Dimension reduction methods based on inverse regression are natural candidates for study as they can identify, not only different mechanisms affecting the mean function, but also those affecting variability. These approaches will be applied to large structure-activity data bases.
|
