Partial correlation matrix and gene cluster

Question

[zhang.jianhai]

Hello Biostar community,

I am using partial correlation methods, i.e. SPACE or GeneNet, to construct gene networks. Can I use the partial correlation matrix for hierarchical clustering so that I can identify the gene clusters? It is common to use the traditional correlation matrix for hierarchical clustering to identify gene clusters. How about the partial correlation matrix?

Regards,

Answer 1

[Jean-Karim Heriche]

You can do clustering using partial correlations. The partial correlation between X and Y controlling for the effect of Z is the correlation between the residuals of the linear regressions of X on Z and Y on Z. One way of seeing it is as the cosine of the angle between the projections of X and Y in a plane perpendicular to the direction of Z.

[zhang.jianhai]

Thank you very much for your reply!