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
176 stars 23 forks source link

TF activities in single-cell: `run_ulm` takes too long #106

Closed nicholaslimwy closed 7 months ago

nicholaslimwy commented 7 months ago

Thank you for developing this package! Unfortunately I've been struggling to construct the univariate linear model with the run_ulm function on my dataset. I assume this is due to the large cell numbers I have (~35k) which failed after 8 hours, but despite downsampling it to 1k cells, the function is still taking just as long. I've seen that an alternative solution is to run the datset using Python, but was wondering if constructing the CollecTRI database, followed by running it on VIPER would be a viable alternative?

PauBadiaM commented 7 months ago

Hi @nicholaslimwy,

May I ask which version of decoupleR are you running? It could be that you are using an older version that is less efficient than the current one. I would try to install it from GitHub and try again:

remotes::install_github('saezlab/decoupleR')

Else yes, the alternative is to switch to python which should take seconds to run on a normal laptop. VIPER, although we saw in our benchmark that it has less predictive performance than ULM, is also a valid alternative, although it can also be slow when its pleiotropy argument is used, for faster (but less accurate) results you can set it to FALSE.

Hope this is helpful!

nicholaslimwy commented 7 months ago

I was running 2.6.0 and have since updated it and it works perfectly! Thank you for your help!