Is there a way to plot the data of the microarray color-coded using some data such as GO term?

Question

[kkam214]

I am trying to plot microarray data on 2D plane by PCA and t-SNE now.

So, I'd like to colorize each function of each gene, but is there any good way?

Multiple GO terms are allocated to each gene, and using a detailed GO term requires tens of thousands of colors for color coding.

If possible, I would like to paint about the rough function of each gene at around 20 colors.

Also, I am pleased that all work is completed on R.

I am sorry to request various requests.

Thank you.

Answer 1

[Michael]

This sounds like an interesting idea, but I cannot think of a good way to color code all possible GO terms in a way that is accessible to human perception. In a scatter-plot of a PCA you might as well end of with no visible structure, that will make it impossible to spot any groups, because most genes are not differentially expressed or have no role under the experimental factors; in the same way most GO terms should not be affected under any given experimental conditions. => only few genes AND only few GO terms are significantly altered at any time. But, if you are not afraid of whipping some visualization up for yourself (which potentially could lead nowhere), you could do the mapping once you focus on the enriched set of GO terms under the process you are studying. You could try the following:

• Get some gene sets, e.g. by DE analysis
• Run GO term enrichment analysis on the sets
• Further reduce the number of relevant terms using e.g. REVIGO tree maps
• assign a color to each gene, based on it being assigned to a child node of the few relevant terms
• chose the color based on semantic similarity (e.g. SimRel) such that similar terms have similar colors, e.g. use the same colors as in REVIGO treemaps

I am sorry, but I don't have an easy code recipe for the last 2 steps. Hope that helps.

[kkam214]

I'm sorry. I did not write any information to tell. I am now working on time series data, and I thought that dimensionally reducing the matrix in the time series direction could be used as approximation of the expression tendency of the whole gene. But I don't think it's the right way to use GO term to color them. So I asked for Biostars.

[Kevin Blighe]

You want to plot the genes in 'pseudo-time'?

[kkam214]

I first learned the concept of 'pseudo-time', I will consult with a co-worker to see if I can write something new in that direction. Thanks