syspremed / explobatch Goto Github PK

If more than 7 batches are used, no colour is assigned for the remaining batches on PCA plots and so they are not plotted.

please use the recommended way of including depending packages while you build your package. as usethis::use_package

https://r-pkgs.org/index.html

I got a very weird error when running exploBATCH.
Error: With R version 3.5 or greater, install Bioconductor packages using BiocManager; see https://bioconductor.org/install

Dose this mean that some dependency packages should be installed from Bioconductor and is not in my session?
Thanks in advance!

Error: With R version 3.5 or greater, install Bioconductor packages using BiocManager; see https://bioconductor.org/install.

Can you please fix it ?

Hello!

SPLS-DA and PLS-DA models built on our data + possible known batch effect show the models can accurately classify samples based to one of two batches (pR2Y = 0.05 & pQ2 = 0.05 based on 20 random permutations of response labels to estimate R2Y and Q2Y significance). However, expBATCH finished after findBATCH and I'm assuming because the findBATCH function reported the batch as not significant. Could you comment on whether we should still remove the batch, even though it's not classified as significant by findBATCH ?

Thanks!
Jenya

Hi,
By default, expBATCH() seem to be using all the cores. Is there any way to limit the number of cores to use.

This as such is quite a useful option. However, I want to use this option because I want to select as many PCs as possible during the batch removal so that I can retain most of the data. However, if all the cores of my servers are used, it quickly runs out of memory beyond 9-10 PCs. I am suspecting the use of many cores is causing the use of huge amount of memory and if I can limit the number of cores (it is ok if it takes longer to run), this problem can be solved.

Any help would be appreciated, thanks.

Hello!

I successfully installed your tool and ran the following example from your manual:

expBATCH(D=Breast,batchCL=batchBreast,Conf=NA, mindim=2, maxdim=9,method="ppcca", SDselect=2)

However, when I tried using my own matrix (D) and batchCL, the following error was given:

Error in Covars[, i] - min(Covars[, i]) : 
  non-numeric argument to binary operator

It appears I did get the complete folder pcaBeforeCorrection but all subsequent folders are empty.

Should I preprocess my matrix prior to passing it to expBATCH? For example, log transformation, scaling, outlier removal, or are these steps all performed when calling expBATCH?

Thanks for any input!
Jenya

Hello,

The dataset I'm working with has 2 batch effects, one caused by the plate used for mass spec and the second by the collection year of the samples. As I cannot pass a matrix for the batch effects, I'm trying to run expBATCH 2x:

Run 1: input the original feature table and a vector of plate batches (set SDSelect = 2)
Run 2: input 2020-06-25_pca_After_PPCCA_correction.txt from Run 1 and a vector of year of collection batches (set SDSelect = 0)

For Run 1, I obtained results for both correctComBat and correctBATCH. However, for Run 2, only results for correctComBat were generated which appear to have remove the 2nd batch effect. correctBATCH returned an empty folder and the error:

Error in tcrossprod(x, y) : 
  requires numeric/complex matrix/vector arguments

Do you know what could be the issue? Thank you for the help!

Below you can find the PCA plots for ComBat & PPCCA correction:

1. Run 1, pcaBeforeCorrection.png (6 plates batch effect)

2a) Run 1, correctComBat pca_After_Combat_Correction.png

2b) Run 1, correctBATCH 2020-06-24_pca_After_PPCCA_correction.png

3. Run 2, ppcaBeforeCorrection.png (4 years of collection batch effect, using 2020-06-24_ppccaCorrectedData.txt as input feature table)

4a) Run 2, correctComBat pca_After_Combat_Correction.png

4b) Run 2, correctBATCH failed

When running expBATCH() with 9 or more batches, pcaBeforeCorrection() will crash and ggplot2 will throw the following error message:

Error: Continuous value supplied to discrete scale

I analyzed the code a little bit, and here is the reason: For 9 or more features, we have cGsub=c(1:ngrps) in line 60 of expBATCH.R, which is an integer vector. We pass this vector as a color argument to aes() in line 23 of pcaBeforeCorrection.R via

fig1<-ggplot(data= sdat1, aes(x, y,colour=cGsub[batchCL])) + ...

ggplot2, however, wants this color argument to be a factor, so we can easily fix this bug by changing the above line of code (line 60 of expBATCH.R) to

cGsub=factor(c(1:ngrps))

This solved the issue for me.