search

quadbio / Pando

Multiome GRN inference.
https://quadbio.github.io/Pando/
MIT License
106 stars 21 forks source link

`coef()` results was changed #46

Closed mkojima123 closed 8 months ago

mkojima123 commented 10 months ago

Hi, Thank you for this package. I using pando in my own data. https://quadbio.github.io/Pando/articles/getting_started.html When I use subset from Signac, coef() result was different. https://stuartlab.org/signac/articles/pbmc_vignette

Strangely enough, when I increase nCount_peaks > 3000, the number of rows in the coef() result increases, even though the number of cells is decreasing.

Is there a reason why this ? thank you.

joschif commented 10 months ago

Hi @mkojima123, unfortunately I don't fully know and also can't really control how Seurats subset function interacts with the SeuratPlus object that Pando creates. I tried to make the object compatible with most Seurat functions, at least the ones that don't change the shape of the arrays. What you describe seems quite strange, and it leads me to believe that subset might not be fully compatible with Pando and I would avoid using it. As a workaround you can just save individual networks created by pando and add then back in after the subset. But anyway, the network would not by really valid for the subsetted object anymore, since inference on it would result in in different coefficients.

mkojima123 commented 9 months ago

Thank you. I'm sorry for the late reply. I understood that it is better not use subset. Thank you for the information.