The issue of calculating loadings after Partial Least Squares (PLS) regression.

[zhangxinru1015]

Hello, I have read your paper and I greatly appreciate you providing such detailed code for our reference. To better understand your work, I tried to replicate the analysis using your code. However, I noticed that when calculating the term loadings, the positively correlated loadings in the code are described as negatively correlated in the paper. Is there a specific reason for this discrepancy?

[justinehansen]

Hi, thanks for your question! The sign of the PLS scores are arbitrary; if I remember correctly, we flipped the scores (x -1) in all analyses simply because I was used to conceptualizing the gradient as going from unimodal to transmodal (instead of from transmodal to unimodal). As a result all loadings also have opposite sign. The sign doesn't add any information, all that matters is that these two poles (red and blue in my case) are opposite.

[zhangxinru1015]

Thank you for your answer! I have another question: After performing gene enrichment analysis and displaying important biological processes using a word cloud, I would like to know which metric the word cloud is based on. I tried using the p-values of the corresponding biological processes or the category scores, but I couldn't match them one by one. I would like to ask for your advice.