Do you know any software to perform a blind clustering analysis on microarray expression data?
I'm not exactly sure how it would be 'blind', but MeV can do various clusterings of gene expression data if you want a nice easy to use program, or in the R language you can use the hclust package. In either case you might want to restrict yourself to only genes differentially expressed, as it can be challenging to cluster 10,000s of gene expression values at the same time. I guess by limiting yourself to differentially expressed genes you are doing an initial dimensionality reduction by partitioning genes into DE and nonDE clusters.
Thanks Jonathan... I will go for MeV...
With blind I mean doing clustering in gene expression starting from normalized gene expression data and without taking into account differentially expressed genes.
From what I have seen most clustering algorithms will suffer if you don't filter genes first. I think the problem of calculating a 30,000 x 30,000 distance matrix first tends to blow up and require more memory than available to most PCs. That said, someone else may know something that can do it.
Hi Paco, which is the platform that you have used for microarrays? There are several libraries in R Bioconductor for Affymetrix...in the Affy site you can also find some other free software.
Another software is GeneSpring but it is commercial. We have an old copy, we can still use it so if you have some particular request you may contact me by email. MeV is good and has several visualization and statistics tools and it is quite platform-independent.
Hope this can help.
Andrea... Grazie ... We are using Affy platforms (exon and expression arrays). In fact I already did the blind clustering with MeV and also with a very classic software called "Cluster"... By the way this is an excelent and robust software that deals almost with everything and very easy to use.
Thanks for your help. Cheers
You do need to reduce your detest using a nonspecific filter. I suggest using the varfFilter or nsFilter in bio conductor. Using these you can first filter out gene probe sets with over all low singnal, ie never expressed. The. Next filter for genes with low variance or standard deviation, this will remove the genes that are expressed at constant levels over your experiments. These filters are non specific in that the variance could be in the replicates and not between treatment groups. Then any clustering method is typically blind. You can use heatmap.2 or similar. For non R clustering I like cluster3.0 and JavaTreeView work nice on any platform, with lots of options for clustering styles. I am more of a fan of multivaiate model based clustering like expectation maximization. Packages like mclust work well for this.
Best
Brian
Heres a link for cluster software
http://bonsai.hgc.jp/~mdehoon/software/cluster/
here is a sample work flow for microarray data
http://www.bioconductor.org/help/workflows/oligo-arrays/#sample-workflow
here is the gene filter function
http://www.bioconductor.org/packages/release/bioc/html/genefilter.html
- Badges
- Science topic
- Similar topics
- Biological Science
- Genetics
- Gene Expression