search

carmonalab / STACAS

R package for semi-supervised single-cell data integration
GNU General Public License v3.0
75 stars 9 forks source link

Deletion of FilterAnchors.STACAS #18

Closed hiuchi closed 2 years ago

hiuchi commented 2 years ago

Thank you for providing a great package.

Now, I am attempting to reproduce the prior study, which uses the STACAS package. However, I can not find FilterAnchors.STACAS in the namespace. Has the FilterAnchors.STACAS function been removed in the latest version?

mass-a commented 2 years ago

Hello @hiuchi!

yes, we have slightly changed the integration pipeline in STACAS v2.0.

Filtering integration anchors is very effective, but we and other users have found that it can be tricky to automatically find a robust threshold for anchor filtering, which required some trial and error to find optimal parameters. For this, in the newest version we opted for anchor downweighting instead of anchor filtering (i.e. lowering the contribution of suboptimal integration anchors, instead of removing them). This seems to be more robust, also because it prevents those cases where too many anchors are removed, making the integration fail.

We are preparing some new demos to illustrate the changes and new features in version 2.0. You can see a couple of new demos HERE and HERE – we'll be adding more, including comparison to other methods.

You can of course reproduce the behavior of previous versions by installing them with a tag or a commit number, e.g.

remotes::install_github("carmonalab/STACAS", ref="1.1.0")

I hope this helps, -m

hiuchi commented 2 years ago

Hi @mass-a

Thank you for your kindly reply. I will try older version or new approaches.

Again, thank you for your sophisticated package!!