The goal in this analysis is to detect groups of genes that demonstrate differential expression between two/more condition groups.
a) t-Test (Supervised Grouping >> Differential Expression >> t-Test): When using this method, genes can be assigned into one of two groups (up-regulated and down-regulated), depending on the definitions of t-test parameters.
b) Wilcoxon\Mann-Whitney two sample rank sum test - (Supervised Grouping >> Differential Expression >> Ranksum test): When using this method, genes can be assigned into one of two groups (up-regulated and down-regulated), depending on the definitions of Ranksum test parameters. This method is nearly as efficient as the t-test on normal distributions of expression values but has a greater efficiency than the t-test on non-normal distributions of expression values.
c) Negative binomial (edgeR) – Negative Binomial distribution test for RNA-seq count data (Supervised Grouping >> Differential Expression >> NB (edgeR)): this method is used to demonstrate differential expression between 2 condition subsets for RNA-seq count data where for each probe i and condition j in the expression matrix the value is a non-negative integer. As part of the test 3 different dispersion options are given: Tagwise – for a large amount of samples (>6), where a different dispersion is calculated for each probe, Common – for small amount of samples (<6), where the same dispersion is given for each probe, Poisson – a special case of NB where dispersion = 0 for all probes. The probes are then assigned into two groups (up-regulated and down-regulated). For further information regarding edgeR please refer to References. Before using edgeR, please make sure you have R software along with the “edgeR” and "limma" packages installed (see R External Application section).
d) SAM - Significance Analysis of Microarray (Supervised Grouping >> Differential Expression >> SAM): this method detects probes that demonstrate differential expression between conditions subsets. You may choose 2 or more subsets (multi-class tests are supported). The probes are then assigned into two groups (up-regulated and down-regulated) if 2 condition groups are tested or into one group of differentially expressed otherwise. SAM uses permutations to get an ’empirical’ estimate for the FDR of the reported differential genes (for details see the References section). Before using SAM, please make sure you have R software along with the “samr” package installed (see R External Application section).
After performing differential expression grouping analysis, a solution visualization tab is added to the main window. It contains the following views:
Information regarding the algorithm, number of groups (can be either 1 or 2), number of un-grouped elements (non-differential), and numerical measures of the groups quality, including:
a) Overall average homogeneity - calculated as the average value of similarity between each element and the center of the group to which it has been assigned, weighted according to the size of the group.
b) Overall average separation – calculated as the average similarity between mean patterns of different groups, weighted according to their sizes.
c) Groups table - contains the number, name (label), size and homogeneity of each group.
Mean Patterns of the groups with error bars (±1 STD).
Upon selecting a group, the corresponding pane is displayed on the right. It contains a list of probes, p-values/q-values, fold-change, probe patterns, expression matrix (heat map) and the chromosomal locations of the genes. Similarity matrices for probes within the cluster as well as for conditions are also displayed in this tab, if the relevant options in the display settings are selected (see the Settings section). If a network file has been loaded (via Data>>Load Network), the sub-graph, induced by the cluster is also displayed in the group pane.
In order to allow comparison between groups and patterns, the displayed expression patterns are automatically standardized to have mean = 0 and STD = 1.
A differential expression solution can be saved using the File >> Export to text..., and reloaded using the Grouping Supervised Grouping >> Differential Expression >> Load Solution.