Saturday, February 25
PS3 Poster Session 3 and Continental Breakfast Sat, Feb 25, 8:00 AM - 9:15 AM
Conference Center AB

Partial Least Squares Regression Analysis Identifies Interleukin-1 Receptor as a Predictor of Airway Neutrophils in Asthma (303465)

Stephane Esnault, University of Wisconsin-Madison 
*Michael David Evans, University of Wisconsin-Madison 

Keywords: PLS regression, multivariate analysis, latent variables, asthma

BACKGROUND: In asthma, the immune response that directs the accumulation of neutrophils in the airway is not well understood, and therapies for neutrophilic asthma are lacking.

OBJECTIVE: To describe the association between sputum gene expression (predictors) and sputum leukocytes (responses).

METHODS: Expression of 16 genes and differentials for 4 leukocytes were measured in sputum samples from 48 subjects with asthma. Because the predictors (genes) are both numerous and highly correlated, multi-response PLS regression models were used to examine these relationships. A 3-component model was selected by cross-validation. Data visualization was aided by correlation matrix plots arranged by angular order of eigenvectors, a novel coupled-seesaw plot of the X and Y loadings of the 3 latent components, and coefficient plots with jackknife CIs for the gene predictors.

RESULTS: Orthogonal latent components regulating eosinophils and neutrophils were identified. The eosinophil-related genes are well known, but the relationship between neutrophils and genes characterizing the interleukin-1 receptor has not been described.

CONCLUSION: Treatments targeting the IL-1 pathway may be beneficial to treat neutrophilic severe asthma. PLS regression provided a successful strategy for handling numerous correlated predictors and identifying an interpretable and biologically relevant underlying structure.