search

saezlab / decoupleR

R package to infer biological activities from omics data using a collection of methods.
https://saezlab.github.io/decoupleR/
GNU General Public License v3.0
183 stars 24 forks source link

Can I still use Findmarkers on dorothea results returned by wmean #51

Closed Chengwei94 closed 2 years ago

Chengwei94 commented 2 years ago

Can I still use findmarkers on the score returned run_wmean? Do I need to scale through scaledata the data first before running it?

Another qn: It seems that there are 11 methods to run in decoupleR. Is there any advise which is the best/what should I run for dorothea or progeny. It seems that the previous preferred method (viper) is no longer preferred?

PauBadiaM commented 2 years ago

Hi @Chengwei94 ,

Thanks for checking out the package!

Can I still use findmarkers on the score returned run_wmean? Do I need to scale through scaledata the data first before running it?

Scaling the data is optional but it helps to better identify differences across cells and yes you can run findmarkers to identify which TFs might be more active in certain groups.

Another qn: It seems that there are 11 methods to run in decoupleR. Is there any advise which is the best/what should I run for dorothea or progeny. It seems that the previous preferred method (viper) is no longer preferred?

In the decoupleR manuscript we observed that methods based on linear models (ulm and mlm) or norm_wmean perform better therefore we recommend them over the rest. In the vignette we use norm_wmean because in this R implementation ulm and mlm take a while to run with more than 1000 samples. If you want to run them faster I would recommend you to check the Python implementation of decoupleR. Hope this was helpful.