I can use the ca object to annotate my query set. The labels are provided by ca.idx2label seem fine. However, there is no further metadata. OTOH, I can use cq to query for nearest neighbors. This works fine. I can get metadata from cq.cell_metadata but the metadata has no celltype labels. I can also get embeddings for cq. This is good too.
However, the indices used by the kNN for ca are different than the one used by cq (they are read in from different files) so I can't map the labels from ca to cq and I can't map the embeddings and metadata from cq to ca.
The metadata and embeddings would be nice so that I could subset the reference data before cell annotation. Conversely, the labels would be nice if I wanted to re-index a new KNN. Right now I can't figure out how to do either.
If it helps I am using you model-v1.1
Thanks
Hong
I can use the ca object to annotate my query set. The labels are provided by ca.idx2label seem fine. However, there is no further metadata. OTOH, I can use cq to query for nearest neighbors. This works fine. I can get metadata from cq.cell_metadata but the metadata has no celltype labels. I can also get embeddings for cq. This is good too.
However, the indices used by the kNN for ca are different than the one used by cq (they are read in from different files) so I can't map the labels from ca to cq and I can't map the embeddings and metadata from cq to ca.
The metadata and embeddings would be nice so that I could subset the reference data before cell annotation. Conversely, the labels would be nice if I wanted to re-index a new KNN. Right now I can't figure out how to do either.
If it helps I am using you model-v1.1
Thanks
Hong