Tool:PCAtools: everything Principal Components Analysis

Question

[Kevin Blighe]

NB - please post questions on Bioconductor Support: https://support.bioconductor.org/

[ direct link to create a post: https://support.bioconductor.org/p/new/post/?tag_val=PCAtools ]

---

PCAtools

Github

• https://github.com/kevinblighe/PCAtools

Bioconductor

• https://bioconductor.org/packages/release/bioc/html/PCAtools.html (release version)
• https://bioconductor.org/packages/devel/bioc/html/PCAtools.html (development version)

vignettes

• https://bioconductor.org/packages/devel/bioc/vignettes/PCAtools/inst/doc/PCAtools.html
• https://github.com/kevinblighe/PCAtools

[Istvan Albert]

One comment I would have is that I think the package should work off simple text files, rather than spending most of the tutorial code that preparing and formats the data to be readable into the package - and this cuts to almost to the philosophy of how Bioconductor packages are usually presented.

Most beginners do not understand what is inside of a pData and in general, one cannot tell unless they first investigate the pData themselves. So when following a tutorial we cannot properly visualize what the input data is.

There should be a movement in the Bioconductor world where every package is demonstrated in terms of "flat" data. The help should start with statements such as: "supppose you have your data in the form":

Name       Cond1         Cond2      Cond3               
GeneA       10             11        20
GeneB       20             22        20
GeneC       30             15        45

Now, what does the package do? How do I read in this file, which everyone has now understood, a file that has an explicit structure and what is the result that we get out of the package At the end of the tutorial, one should list examples of how to "make" the same flat format from a more complex one. But always starting with pData (which itself is a remnant from the dinosaur era) hinders people who would use the package but do not want to need to use a pData format. In addition, simplicity begets simplicity, code and use cases would end up simpler and more generic if the input data were simple.

[Istvan Albert]

Oh, I got caught up with "opinions" and forgot the most important message - great library - we were investigating options for PCA plotting this very morning - very timely and thanks for the work!

[Kevin Blighe]

No worries. Indeed, there's nothing else out there for creating bi-plots, etc in R. I'm aware of a commercial solution for PCA that's being used, but it's expensive. Yes, the vignettes could be more user friendly. I start with random data matrices or pre-existing datasets. You may want to take up the other issues with Martin Morgan ;)

[1769mkc]

it looks like lot of people have started using it ..

[Kevin Blighe]

Are you looking at download metrics? It is still not officially released yet

[prasundutta87]

Great package..but why does it want only R 3.6..It's not even officially released.

[Kevin Blighe]

That is stipulated by Bioconductor. New official release packages require the next version of R and Bioconductor. This is ultimately positive because it ensures that users stay up to date.

If you install via devtools::install_github('kevinblighe/PCAtools'), it should install and run okay.

Sorry about that.

[prasundutta87]

Understandable..Thanks a lot Kevin..

[Kevin Blighe]

I should add that, upon official release, the R 3.6 requirement will actually be removed. For now, the development version has passed all checks and can be used.

[1769mkc]

"Error in biplot(p, showLoadings = FALSE, lab = NULL) :   unused argument (showLoadings = FALSE)"  why this argument is not working?

[Kevin Blighe]

Hey mate, you may need to update the version of PCAtools. Which version are you using?

[1769mkc]

@Kevin this one PCAtools’ version 1.0.0 and this your 2.2.0 current version ,,I update all my R library quite frequently shouldn't it be updated automatically? I tried updating its not getting updated

[Kevin Blighe]

If you also use a previous version of R, then that will limit which version of PCAtools that you can use

[1769mkc]

i updated using your developer version and it worked

[1769mkc]

"showLoadings" this shows the magnitude of contribution of those features to the pca in terms of length of the arrow if i understand and what does the direction signifies

[rodd]

Wowza! The plots I generated using PCAtools have already upgraded my future manuscripts I think! Thank you, Kevin Blighe et al.!!

Can PCAtools plot pre-computed principal components (e.g., from genetic data, from plink)? 👀

[1769mkc]

this might be helpful PCA