Time-course microarray experiment is capable of capturing the dynamic profile of genomic response to treatment factors. The profile contains valuable information for researchers to identify possible genetic factors that lead to different clinical outcomes, which can help directing future investigation. We developed a general statistical method to extract gene-specific temporal patterns to the interaction of multiple treatment factors.
· First rigorous statistical method for analyzing time course multifactor data
· Enabling the identification of gene specific response timing and pattern to experimental factors.
We developed a statistical inference method for time course multifactor analysis. This method is applicable to both longitudinal and cross-sectional microarray data. It can handle both balanced and unbalanced experimental design, a frequent situation in observational studies.
1. Estimate the response feature (γ1, γ2).
2. Project time course data to the estimated response feature space.
3. Classify genes into the following five classes using non-parametric ANOVA (C1-C4 contain useful information. An example is shown in the figure).
4. Cluster genes by their projection vectors within each class.
For a detailed description, please check out the slides! Oct 6 2008 talk, Dec 8 2008 talk
· Longitudinal Multifactor Analysis
· Cross-sectional Multifactor Analysis
Baiyu Zhou, Weihong Xu, Wenzhong Xiao