ANOVA: analysis of variance, used to determine whether the factors (such as different samples, treatments, or time points) considered in the model have significant effects on gene expression levels. The fold-changes (or logratios) can be calculated as various parameters in the model, and error bars and p-values are also estimated to provide the statistical significance of the analysis.
Click here to view the example.
FDR: false discovery rate, used for multiple test adjustments. Briefly speaking, it estimates the portion/percentage of false positives in a list of candidate genes. They are calculated from the raw p-values for each gene by some statistical methods (such as ANOVA or Mixed Model).
Click here to view the example.
 Clustering: Genes can be grouped together according to the similarity of their expression levels across several experimental conditions. It can establish some type of associations among genes, and be very helpful in inferring gene functions.
Click here to view the example.
 PCA: principal component analysis, a method of “compressing” the variability of gene expression profiles (usually involving tens of thousand of genes) into a small number of “principal axis/component” (usually 2 or 3) so that the differences among samples/treatments can be easily visualized in a 2 or 3 dimensional plot.
Click here to view the example.
Pathway analysis: Some genes on the array may be involved in some well-established biological pathways. By look at the expression levels of genes in the pathway collectively, we can answer the question of whether a particular pathway is significantly affected by different treatment. P-values are calculated to quantify the significance of changes.
Click here to view the example.
|
