search

satijalab / seurat

R toolkit for single cell genomics
http://www.satijalab.org/seurat
Other
2.3k stars 918 forks source link

FindTransferAnchors() and TransferData() "dims =" argument #3400

Closed danielcgingerich closed 4 years ago

danielcgingerich commented 4 years ago

I am using FindTransferAnchors() and TransferData() to annotate cell types using a reference dataset. My reference dataset is much larger than my query dataset. As such, when performing jackstraw, each dataset has a different number of significant principal components. So, for the "dims =" argument of the FindTransferAnchors and TransferData functions, since the reference and query each have a different number of significant PC's, which one should I put in this argument? Is "dims =" referring to the dimensions of the query object? R documentation does not specify.

satijalab commented 4 years ago

For transfer anchors, we would suggest the dimensionality of the reference dataset.

ElyasMo commented 1 year ago

I am sorry but it was unclear to me.

The process includes two steps which are FindTransferAnchors and TransferData. They both have the dims parameter. Do you mean for both of them we should prioritise reference over query set for the number of dimensions?

JerryZhang-1222 commented 8 months ago

I am sorry but it was unclear to me.

The process includes two steps which are FindTransferAnchors and TransferData. They both have the dims parameter. Do you mean for both of them we should prioritise reference over query set for the number of dimensions?

Hi ElyasMo,

According to the vignettes from Signac, I think for FindTransferAnchors we should use the reference dim while for the TransferData we should use the query dim. You can see that in the vignettes of Signac. FindTransferAnchors uses 1:40while in the TransferData, dim = 2:30. Given that usually, we won't include the first dim in ATAC-seq, I guess in the TransferData we should use the query's dim.