Heatmap clustering method, Euclidean or Pearson correlation

[KJ Lim]

Dear guru,

The analyze_diff_expr.pl executes the Trinity-included PtR script to plot a heatmap using euclidean clustering method for gene and sample distance.

  PtR -m diffExpr.matrix --log2 --heatmap --min_colSums 0 --min_rowSums 0 --gene_dist euclidean --sample_dist euclidean --sample_cor_matrix --center_rows --save

However, the default clustering method of PtR script for gene and sample distance is Pearson correlation.

May I ask, which clustering method (Euclidean or Pearson correlation) is suitable for RNA-Seq different expression data?

Thank you for your time.

Best regards,
Lim

[Brian Haas]

It's entirely up to you.  I often use Euclidean.  You can run PtR yourself with the different options.  Use -h for all options, as there's a bunch of different clustering methods to choose from.

--
You received this message because you are subscribed to the Google Groups "trinityrnaseq-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to trinityrnaseq-u...@googlegroups.com.
To post to this group, send email to trinityrn...@googlegroups.com.
Visit this group at https://groups.google.com/group/trinityrnaseq-users.
For more options, visit https://groups.google.com/d/optout.

--

--
Brian J. Haas
The Broad Institute
http://broadinstitute.org/~bhaas

 

[Brian Haas]

It probably depends on whether the data are centered or not, and also depends on how you want to cluster (including intensity or based on correlations)

On Wed, Feb 24, 2016 at 2:16 PM, KJ Lim <jinke...@gmail.com> wrote:

Dear Brian,

Thanks for your prompt replied. 

I recently read a post suggesting that euclidean clustering method may not good for RNA-Seq. I am confuse. That’s why I am asking around the Gurus here. 

Cheers. 

Lim

[KJ Lim]

Thanks for the explanation. 

This question may out of topic a bit. 

I want to see profile of a set of genes from RNA-Seq differential expressed list. 

Can I use the PtR to plot a heatmap with euclidean clustering method after subsetting them from RNA-Seq differential expressed matrix? 

Thanks for your time.

Cheers.

Lim

[Brian Haas]

Absolutely.  I often do this.  Just be sure to get the top line for the column headers, and then you can use any subset of the rest of the matrix.

best,

~b

[KJ Lim]

Dear Brain, 

Good to know, good to know. Thanks for your time :)

Cheers. 

Lim