bnprks
commented
10 hours ago Hi @wuwei77lx,
Thanks for your question. I would suggest starting by reading the section "scATAC-seq reference-based cell type annotation" from the methods of our paper. We cover there the specific tools and functions used, along with some non-default parameter settings used in those calls. For the specific task of cell annotation, we did not use Signac, and that section of the methods describes how we used Azimuth, Seurat, and a healthy multiome PBMC reference to transfer labels to scATAC-seq data. (Although we took some suggestions on how to use the Seurat functions on scATAC-seq data from a Signac vignette, the functions themselves were from Seurat)
Different computational methods can perform better or worse depending on the specific dataset, so I think the most direct way to evaluate annotation quality is by looking at the annotations themselves. We've provided our cell annotations via metadata files in this repository with the specific per-cell annotation assignments, along with the original fragment file data available via GEO accession GSE174072. If you want to evaluate our cell annotations, I'd encourage you to use those files as your starting point.
Best wishes, Ben
Hi,I am using the scATAC seq data from your article 'Multi-omic profiling reveals widespread dysregulation of innate immunity and hematopoiesis in COVID-19'. The effect of using Signac software for cell annotation is very poor. I would like to ask if you have used Signac for cell annotation, or can you provide detailed information（code）on ArchR software annotation? thank you!