RP: Rank Products

(Breitling, Rainer et al, 2004)

Rank Products is a novel test for determining differential expressed genes with multiple replicates. This analysis differs from many other techniques in that it does not apply a sophisticated statistical model, but rather from the calculation of rank products, a faster and simpler method.

Additionally, Rank Products is useful in highly noisy data and can significantly reduce the number of replicate experiments required to obtain reliable results.

Rank Product Initilalization Dialog

MeV’s RP currently supports 3 experimental designs.

1. One-class, typically run on two-color data, this design determines genes that are significantly up or down-regulated within the included group. To exclude a sample from the analysis, uncheck the box next to that sample’s name in the left pane of the one-class screen.
2. Two-class unpaired, where samples fall in one of two groups, and the subjects are different between the two groups (analogous to a between subjects t-test). The initialization dialog box is similar to the t-test dialog.
   The user inputs the group memberships of the samples in the top panel. In the two-class design, genes will be considered to be “positive significant” if their rank product in group B is significantly higher than in group A. They will be considered “negative significant” if the rank product of group A significantly exceeds that of group B.
3. Two-class paired, in which samples are not only assigned to two groups, but there is also a one-to-one pairing between a member of group A and a corresponding member of group B (e.g., gene expression measurements on a group of subjects, where measurements are taken before (Group A) and after (Group B) drug treatment on each subject).

Targeted Genes

As in most other modules, MeV offers two forms of sample selection: the individual button selectionthe cluster selection tabs to assign your samples to the analysis. Samples left unassigned or unchecked will be ignored in the analysis.

P-Value/ False Discovery Parameters

For determination of significance levels, specify the number of random permutations you want RP to run. If setting a significance cut-off using p-values, enter the alpha value you wish to set as the cut-off point. If determining significance by false discovery rate, check the box next to either the number or percentage of false positives and input the corresponding value.

Targeted Genes

Check the radio button for determining in the analysis significantly up-regulate genes, down-regulated genes or both.

Hierarchical Clustering

Hierarchical Clustering

To have RP construct hierarchical trees for your results, check the corresponding check box. Select whether this feature is to be applied to significant genes or significant and non-significant genes. This process may add significantly to the computation time.

The RP module outputs standard viewers and tables for MeV’s statistics analyses.