GV Exercise.6
Perform clustering and find genes with similar profiles
[ Main_Page | Genevestigator_training | Analyze_public_microarray_data_using_Genevestigator | GV Exercise.5 |
| GV Exercise.7 ]
last edit: October 31, 2014
Contents
Aim
Find probes similarly regulated from the full compendium. Identify conditions where samples and probes show reciprocal similarity. Create biomarker lists and find co-expressed genes.
Find ABA induced markers from Ath.all
- Select the Ath.all' Sample Selection and open the Gene search tool with perturbations
- collapse all then expand 'hormone' and find 'ABA (1h)' (vs mock treated seedlings (1h)
try it first
- check sensitivity to find 25 biomarkers with a min Target log-ratio of |2| (4x) (0%tolerance) and run the tool
try it first
- save the 25 biomarkers to a new Gene Selection and name it ABA_1h-25URLR2
try it first
- also save the results as excel table (ABA_1h-25URLR2.xls) in case you wish to run them in a functional enrichment program.
- try finding DR genes with the same parameters
Perform hierarchical clustering of the ABA markers and all Ath perturbations
- Select the ABA_1h-25_URLR2 Gene selection and the Ath.all Sample Selection to cluster probes and conditions and find similarity.
- start the hierarchical clustering tool, set it to perturbation with genes and conditions checked and euclidian distance, and hit Run.
- scroll down to the bottom of the obtained heatmap
try it first
Notre that you can cluster by samples AND by genes in order to partition your graph with similar entities close to each other. This is useful to find genes with similar expression patterns across a subset of conditions and thereby find similarity between conditions that otherwise might not be obviously related.
Apply bi-clustering and identify most homologous sub-clusters
As such a long heatmap is not very handy, we apply bi-clustering to identify subsets of samples and perturbations that show most similarity. This process is unsupervised and will find many independent solutions to the problem. You must review the solutions and select one that matches your expectations. Bi-clustering find groups of probes that are co-regulated under a subset of conditions. In many cases, this will reveal similar transcriptional control of the genes through the presence of identical cis elements in their promoters. Bi-clusters may correspond to different groups of genes sharing the same combination of cis-elements within each group and various degrees of overlap between groups (clusters).
bi-clustering can be done for up- or down- or both differential expression directions
The bi-clustering analysis often finds too many solutions. In such case, set the minimal count of probes and samples higher in order to make a more stringent run and retry
- set the stringency to min 10 conditions (factors) and min 20 probes
try it first
Try bi-clustering with the HCC list vs liver and see what Anatomy parts are represented
| first two bi-clusters | |
|---|---|
 |
 |
 |
From the results, it is clear that the hormone ABA has very similar gene induction profiles to those induced by stress (salt, drought).
bi-clustering results can be used to create NEW Sample Selections AND_OR NEW Gene Selections (mouse select the wished area and create set)
Download the exercise files
Try it by yourself before expanding on the right!
genevestigator workspace
- download ex6.gv4 and open it from within genevestigator File Load Workspace file link
References:
[ Main_Page | Genevestigator_training | Analyze_public_microarray_data_using_Genevestigator | GV Exercise.5 |
| GV Exercise.7 ]
