changing parameters of pca plot to show results

Question

[Aldo]

hello everyone, I'd like to change the colors of my pca plot to better interpret the results. this is my matrix (23240 elements) with normalized data (I reduced data for brevity)

 WTYETMM
                     sample1     sample2
Sox17             18.858747    17.214335
Gm6123            25.144996    28.690559
Mrpl15           544.808244   558.509546
Lypla1           192.778302   148.234554
Tcea1            449.466801   452.354478
Gm6104            18.858747    16.257983
Atp6v1h          320.598698   341.417651
Rb1cc1           296.501410   309.858036
Pcmtd1           550.046785   585.287401
Gm38372           61.814782    61.206526

first of all i normalized the data using the tmm method, then I ran this script:

library(ggfortify)
prcomp= prcomp(WTYETMM, scale.=TRUE)
autoplot(prcomp, loadings = TRUE,
         data = WTYETMM) 

now i show you the pca i got, but i'd like to show the names of the two samples and color them differently. I appreciate your help. please do not send me links, I would appreciate your comment

[ATpoint]

There is a lot going on here. First, which kind of data is that? The magnitude of values suggests it is not log-scale, so that would be the first thing to fix. Scaling counts is not really meaningful in most cases since you want to put emphasis on rows that change much between columns and scaling dampens this effect a lot. Also, one would commonly do a PCA row-wise, not col-wise as you do it, so pca(t(input)) unless the col-wise operation is really intended. Selection of genes that are variable would probably make sense as well as 23240 genes is a lot and only few probably contribute to separation of samples.

The mentioned PCAtools does all that for you, even though it is not hard to code that up yourself, but why not using existing packages that do it already for convenience and also provide plotting routines.

[Kevin Blighe]

A quick comment: you may try my R / Bioconductor package: PCAtools