Sample Collection and Microarray Processing
The number of samples processed with Affymetrix U133 plus v2 GeneChip microarrays (including more than 54,000 probe sets corresponding to ~25,000 genes in human) for each of the four tissue types, including fat, muscle, skin and blood, is listed in the table below.
The gene expression data of each tissue type were normalized separately by the software RMAexpress.
Time-series linear mixed model is applied to identify genes differentially expressed between patients and controls within each tissue type. The regression formula is
Expression = Indicator * (1 + Time + Time² + Time³) + Individual random effect
Where indicator=1 if the sample comes from patients and 0 if the sample comes from controls. Time is the time when sample was collected since the burn injury, measured in hours. The pvalues of these four factors in the regression formula are converted into False Discovery Rates (FDR) using the qvalue package developed by Storey lab.
Max Fold Change (maxFC) is defined as follows. Divide the time since injury into four bins: 0-4 days, 4-10 days, 10-22 days, and after 22 days. For each bin, calculate the averaged fold change across all the samples falling in the bin. The one with the largest absolute fold change is maxFC.
Significance Criteria for genes
False Discovery Rate (FDR) <0.05 || Max Fold Change (maxFC)| >1.5
K-mean clustering of expression values of each tissue scaled with respect to its control level.
Note that each tissue is analyzed separately.