clustering RNA seq data

[gfmgfm]

Hello,

I have fpkm values obtained from cufflinks-cuffdiff package.
I would like to cluster now the data for the diff. expressed genes (as done oftenly in microarrays).
Any suggestions on how to do the clustering?

Thanks

[Simon Anders]

Have a look at the vignette of our DESeq package, which contains some suggetsions for clustering.

[gfmgfm]

Thanks a lot!
My data is in fpkm (after cufflinks and cuffdiff). Is this an appropriate input to DESeq?
With such values - can I use DESeq to check for differential expression and to cluster the genes?

[Simon Anders]

No, please don't See the many threads here discussing this.

[gfmgfm]

OK, thanks a lot.
Is there any way to go from cufflinks FPKM to DESeq?

[priya]

Sample clusteringg of RNA-Seq data

Hi,
I have 20 samples of RNA-seq data with fpkm values without replicates. I would like to do hierrachical clustering to look how the smaples were clustered. I have come across MBcluster.seq package in R which do the cluster analysis with Count data and in replication situation.
I am wondering with following questions:
1. Can I do sample clustering across various samples to loook how they are related?
2. Is it good way to do clusteing with FPKM values instead of actual reads(Count values)?
3 Is there any R package which deal with clustering without no replicate situation?

Any ideas is greatly helpful.
Thank you

[Simon Anders]

1. Sure. BTW, besides hierarchical clustering, also try PCA or MDS. Especially MDS is often quite helpful.

2. You can try to take the logarithm of the FPKM values and see how far you get with that. It is important that your data is homoskedastic to get good results. To check, plot the variance of the genes across samples against the ranks of their means. You should notice no (or only weak) dependence of variance on mean.

3. There is no need for replicates if all you want to do is clustering.