search

stephenslab / susie-paper

Experiments on SuSiE ("sum of single effects") regression.
https://stephenslab.github.io/susie-paper
MIT License
5 stars 3 forks source link

QTL experiment #14

Open garyzhubc opened 2 months ago

garyzhubc commented 2 months ago

Just trying to understand the QTL experiment a bit more.

The QTL dataset is unavailable.

From the paper it seems unclear whether p=600 or p=77345. Did all SNPs run in one susie model or 77345 separate susie models?

I'm guessing it's p=600 of 77345 susie models. The outcome variable is x_d = "intron (d) with some bp up and downstream" and y_d = "intron ratio for intron (d)" for d = 1, ..., 77345.

Am I correct?

If so, what are the traits? In the code it's trait=AC. What does this mean? Is it talking about the outcome?

garyzhubc commented 2 months ago

If what I read from the paper is correct, then the intron ratio is a 77345 by 1 vector. But the paper claims it to be a matrix. I'm confused.

garyzhubc commented 2 months ago

By "regressing out PCs", does the author mean subtracting the y_d's predicted by PCs from y_d's or using the y_d predicted by the PCs directly instead of using y_d themselves?