Modeling and Clustering of Time-Course Data

• Activity Number: 284
• Type: Topic Contributed
• Date/Time: Tuesday, August 3, 2010 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Alejandro Murua*+
• Companies: Université de Montréal
• Keywords: clustering ; time-course ; smoothing ; nonlinear differential equation ; gene-expression ; finite mixture

Abstract

We present a methodology to simultaneously model and cluster independent replicates of time-course evolution data. We assume that the time course signals arise from non-linear second order differential equations (2NDEs). The signals are smoothed to better estimate the coefficients of these equations. The smoothed time-courses are obtained by penalizing large deviations from these equations. The 2NDE coefficients are used instead of the original signals to cluster the data. The data are embedded into a finite mixture model that allows the simultaneous estimation of parameters and clustering. We have applied this methodology to the analysis and clustering of gene-expressions related to the development of the retina in mice.