search

sararselitsky / FastPG

Fast phenograph, CyTOF
Other
25 stars 6 forks source link

Is it possible to adjust the resolution? #20

Open ahorisberger opened 2 years ago

ahorisberger commented 2 years ago

Hi,

Thank you for this great tool, it does allow much faster clustering of large datasets. Using the default parameter, I get a very high number of clusters and I was wondering if there was a way to adjust the resolution while running the louvain clustering function, as it is done in Seurat?

Thank you very much for your help

sararselitsky commented 2 years ago

Thanks for your request! Stuart, can that be a parameter that's altered?

On Tue, Jul 26, 2022 at 1:15 PM ahorisberger @.***> wrote:

Hi,

Thank you for this great tool, it does allow much faster clustering of large datasets. Using the default parameter, I get a very high number of clusters and I was wondering if there was a way to adjust the resolution while running the louvain clustering function, as it is done in Seurat?

Thank you very much for your help

— Reply to this email directly, view it on GitHub https://github.com/sararselitsky/FastPG/issues/20, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACQD6TZ6MBSCGOXM6BOENLDVWAMKJANCNFSM54WWOYXA . You are receiving this because you are subscribed to this thread.Message ID: @.***>

jefferys commented 2 years ago

I do not believe that is directly possible with any of the exposed parameters from the louvain software (Grappolo) we are wrapping. Some of them may affect convergence indirectly. If some of the clusters have way fewer members than others, unless you are seeking rare types, you could drop those due to low support. You could also try reducing the variability in some features, or try clustering with some features left out. A TSNE plot might help you identify clusters with more obvious separation/interpretations.

ahorisberger commented 2 years ago

Thank you very much for you answer, I'll try these options.

sararselitsky commented 2 years ago

Thanks, Stuart! Mahantesh, can the resolution parameters be easily adjusted for the Louvain step?

On Tue, Jul 26, 2022 at 2:59 PM ahorisberger @.***> wrote:

Thank you very much for you answer, I'll try these options.

— Reply to this email directly, view it on GitHub https://github.com/sararselitsky/FastPG/issues/20#issuecomment-1195866085, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACQD6T7HAEKMPPFUCTNKL3TVWAYPZANCNFSM54WWOYXA . You are receiving this because you commented.Message ID: @.***>