-
Notifications
You must be signed in to change notification settings - Fork 4
Open
Description
Hi, thanks for your comprehensive workflow. But I have several question.
I use the intron excision ratios in a group of samples as the "expression file", and use this pipline to construct the predictive model. However, when I was processing the filter_db procedure, there were only several models retained (zscore_pval < 0.05),
- I wonder if it was because of using "ratio" as the input?
- Are "models.db" and "models_filtered_signif.db" can be both use to fit the spredixcan?
Metadata
Metadata
Assignees
Labels
No labels