Xin Gao

York University

Constructing Gene Networks Using Hidden Markov Models
for Time Course Microarray Data     

Gene-Gene interaction has been a very important issue in
genetic field as it provides the crucial understanding to the gene
network underlying all different biological mechanisms. Time
course microarray data provides a good platform to reveal how the
interaction among different genes evolves along the time. By
considering the hidden Markov process underlying the time series,
we model the observed gene expression measurements to be dependent
on the unobserved hidden states. The marginal distribution of the
measurement conditional on the expression state and the mixing
proportion of the DE versus NDE genes was estimated marginally at
each time point using the nonparametric empirical Bayes method. A
hypothesis test for interaction is assessed by bootstrapping
method to resample the hidden processes. The control of false 
discovery rate in forming the edges of the gene network is also