Question: Regression Coefficients Matrix

0

vrrani •

**10**wrote:I have computed regression coefficients between mRNA expression values (Y) and methylation features (X).

So, I have a rectangular matrix with regression coefficients signifying the influence of Y features on X.

My query is, can I perform Euclidean measurement on the regression coefficients/beta matrix, in order to get a square matrix?

Any help is highly appreciated.

Thanks in advance.

Regards,

Reddy Rani V.

Hi Reddy - interesting work. It would be interesting to see the distribution of the beta coefficients via a histogram. You don't have to post here, but which distribution does it follow when you generate it? Also, are you not just interested in the coefficients that have a statistically significant p-value?

42kDear Kevin,

Thanks for the reply.

Yes, I am interested in beta coefficients. But, for example, if I would like to study similar behavior of mRNA genes alone. Can I compute Euclidean measurement of the beta matrix to get an N x N matrix from my M x N beta matrix?

Regards,

Reddy rani V.

10You can do that but I am not sure where you are going with that line of analysis. It is certainly an interesting start to what is a multi-omics analysis, though.

42kDear Dr. Kevin,

I am interested to create a network and study the associations of the genes. So, can I still consider gene-gene associations of mRNA features as multi-omics in this case as I started it with regression coefficients?

Thanks a bunch.

Regards, Reddy Rani V.

10Oh, you are only doing a network analysis? You can do anything that you want with your own data, but you have to be careful about what it means, biologically. A 'network' constructed from Euclidean distances is essentially a different representation of a dendrogram / tree. You may have to think for a while about what it means when you construct a dendrogram / network using beta coefficients.

42kDear Dr. Kevin,

Thanks for the suggestions, it is really helpful for our research.

Regards,

Reddy Rani V.

10