Hi , I have a few questions about use cases for the pigengene package.

1. For our dataset, I provided both a training dataset and a test dataset, and the one.step.pigengene function did not appear to take the test data into account? The output was exactly the same with or without the test data parameters provided (testD and testLabels). My question is, does the test data have to be same dimension or similar sample size as the training set in order for the function to provide the accuracy of trees on the test data?

2. If we had already generated our own WGCNA with our own module eigengenes calculated for each sample, could we use that instance instead of running the wgcna.one.step function or the one.step.pigengene function to regenerate a WGCNA instance?

Thank you for your time and help!

Question 1:

a) Please use the devel Version 1.29.4 in which I resolved some bugs.

b) You should get heatmaps for test data. Also, some performance statistics on train and test data are shown on the decision trees and are returned in the output: e.g., one.step.pigengene(...)$5treeRes$c5Trees[[1]]$info

Question 2:

Most likely yes, depending on what exactly you want to do. As explained in the vignette, the next step in the pipeline is to run the compute.pigengene() function, which requires data, labels, modules (which you can obtain from your own WGCNA analysis ). This gives you a pigengene object, which includes eigengenes. If you have already your eigengenes computed, you can still run learn.bn(pigengene=NULL, Data=<your eigengenes>, ...). However, the make.decision.tree() function does require an object of class Pigengene as an input.