Correlating Hormones to Metabolites (Omics Data)

[#1]

Hello all,

I am trying to find correlations between the expression of hormones and metabolites during different phases of growth from my company's omics data. I have structured my data with 4 columns: Type (hormone or metabolite), Name (of said compound), Phase (A,B,C or D), and Normalized Phase Averages (this is the raw data which is the level of expression of each compound. I have normalized for each type and averaged by phase). I can restructure the data if necessary. 

I was thinking of making a graph where I have hormones on the Y axis and Metabolites on the X axis, looking like a multivariate graph with correlations / heat map. I want to make one of these graphs for each of the phases of growth and then maybe one for the overall correlations of all phases at once. 

I've tried making this in Graph Builder and using Analyze > Multivariate, neither of which have worked. Multivariate won't let me due to the categorical nature of the first three columns.

Please help me, what should I do?

[#2]

Hi,

It is challenging to give you precise feedback without having an example (mock data will do) of your data layout. Still, the first thing you should consider is to split your Metabolites and Hormone data into separate columns. Indeed, JMP needs your X and Y data in separate columns.

Second, you could build a GraphBuilder plot with your Hormone data as X and Metabolite data as Y, and the Phase as WRAP factor.

You can then use the Correlation Ellipse + Dot Plot displays to visualize the relationship between Hormones and Metabolites. Finally, on the left interactive menu next to the plot area, you can select to display the r square and correlation for each panel.

Let us know if this works.

Sorry if this explanation is trivial

Best

TS

[#3]

Hi @Kotton_King ,

It sounds like you have your data stacked (values in a single column). You can try to split the data so that each hormone and metabolite have their own column of values. Then Multivariate will work with these new columns. Also look into Cluster Variables. It will give you a heatmap of the correlations as well as cluster them.

Cheers,