Statistical Modeling of OMICS data Min Zhang, M.D., Ph.D. Department of Statistics Purdue University
OMICS Data Genomics (SNP) Glycoproteomics Lipdomics Metabolomics
Outline Statistical Methods for Identifying Biomarkers Metabolomics Align GCxGC-MS Data Other Projects
Statistical Methods for Identifying Biomarkers Classical Methods Bayesian Variable Selection Regularized Variable Selection
Feasible Easy to implement Incorporate a large number of factors
Regularized Variable Selection Fast Do not need to calculate inverse of any matrix As fast as repeating an univariate association study serveral times
Regularized Variable Selection Fruitful Effective and efficient for variable selection OMICS data in CCE Genome-wide association study Epistasis Gene-gene interactions eQTL mapping
Regularized Variable Selection More Details Will be presented by Yanzhu Lin in the future
Alignment of GCxGC-MS Data The Two-Dimensional Correlation Optimized Warping (2D-COW) Algorithm
The 2-D COW Algorithm
Applying the 1-D alignment parameters simultaneously to warp the chromatogram A Toy Example
Align Homogeneous Images (TIC)
Align Homogeneous Images (SIC)
Align Heterogeneous Images (SIC)
Align Heterogeneous Images (TIC)
Align Chromatograms from Serum Samples
Other Projects Identify Differentially Expressed Features in GCxGC-MS Data Integration of OMICS data Other Clinical Data More …
Summary Regularized Variable Selection Method for Identifying Biomarkers The 2D-COW Algorithm for Aligning GCxGC- MS Data It can also be used to align LCxLC, LCxGC, GCxGC, LCxCE, and CExCE data In Progress Identify Differentially Expressed Features in GCxGC-MS Data
Acknowledgements Dabao Zhang Yanzhu Lin Fred Regnier Xiaodong Huang Dan Raftery