jkobject

/

scPRINT

like 2

Hello!
I have some questions about "gene_embeddings.parquet". This file has 44756 genes x 1024 columns.

1. Do the 1024 columns (dimensions) include all three info mentioned in the paper? Gene itself, location in the genome, and gene expression in the cell?
2. Is this from a pretrained model? I saw the annotations like large, medium, and small.
3. Did you take an average of gene embeddings across all cells? Or randomly choose some number of cells?