Time Course Classification of Treatment Response for Psoriatic Patients

• Activity Number: 318
• Type: Contributed
• Date/Time: Tuesday, August 6, 2013 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistical Learning and Data Mining
• Author(s): Joel Correa Da Rosa*+ and James G. Krueger and Mayte Suarez-Farinas
• Companies: Rockefeller University and Rockefeller University and Rockefeller University
• Keywords: Microarray Analysis ; Time Course Classification ; Threshold Gradient Descent Regularization ; Gene Set Variation Analysis

Abstract

In psoriasis, attempts have been made to predict response to treatment from either baseline skin or blood samples with limited success. In this study we analyze time course gene expression profiles of 24 psoriatic patients at baseline and after 1 and 12 weeks of treatment with Ustekinumab 90mg. Our main goal was to identify genes that play an important role for discriminating responders from non-responders. Response was defined by PASI75, a binary outcome associated to the status of psoriasis. The methodology used two different classification methods T-PAM(Time Prediction of Microarray Analysis) and TGDR. The T-PAM provides LDA-projections of time course gene expressions that are used in the classical PAM and TGDR for classification purposes. Additionally, we explored building a classifier based on a pathway signature rather than gene-by gene-signatures. For this purpose we generate pathway-scores for canonical pathways using Gene Set Variation Analysis (GSVA) and used them to build pathway-based signatures. Using the complete time course we were able to define gene signatures that predict response with high accuracy as measured by cross-validation experiments.