|
Activity Number:
|
433
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 5, 2009 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
ENAR
|
| Abstract - #305672 |
|
Title:
|
Identification of Disease Biomarkers Using Logic Forest
|
|
Author(s):
|
Bethany Wolf*+ and Elizabeth Slate and Elizabeth Hill
|
|
Companies:
|
Medical University of South Carolina and Medical University of South Carolina and Medical University of South Carolina
|
|
Address:
|
Suite 303, Charleston, SC, 29425,
|
|
Keywords:
|
biomarkers ; ensemble methods ; logic regression
|
|
Abstract:
|
Screening tools facilitating early disease diagnosis or identification of persons at risk may reduce disease-related morbidity and mortality. Increased sensitivity and specificity may result from diagnostic tests using multiple biomarkers. Logic regression (LR), a recently developed multivariable classifier predicting binary outcomes from logical combinations of binary variables, yields interpretable models of complex biologic interactions. However LR's performance degrades with noisy data. We extend LR methodology to an ensemble of logic trees called Logic Forest (LF). We conduct a simulation study comparing the ability of LR and LF to identify interactions among variables predictive of disease status. Our findings indicate LF is superior to LR in identifying important predictors. LF is a flexible statistical tool for identifying variable interactions associated with disease.
|
