Basis matrix is different for the same snRNA-seq but different spatial samples

[Qirongmao97]

Hi, thanks for presenting CARD.

One question regarding the basis matrix construction, I've used the same snRNA-seq to deconvolute different spatial samples, but the gene list in the basis matrix B for different spatial samples is different. I was wondering if that makes sense cause I thought the basis matrix construction is based on the single-cell data, and if we do not change the single-cell data, the basis matrix will not change.

Thanks!

[YMalab]

Hi @Qirongmao97,

Thank you for your interest in our CARD package.

Regarding your question, because the spatial samples differ, even when we apply the same filtering criteria to remove lowly expressed genes in spatial samples, the resulting gene list will vary. CARD deconvolution considers the genes shared between the scRNA-seq reference and the spatial transcriptomics data. Therefore, the set of genes in the basis matrix used may differ due to the specific common genes shared with the spatial transcriptomics data.

Best, Ying

[Qirongmao97]

Thanks for your reply!

In our case, we do notice that some genes are removed in some of our samples even though they are highly expressed and regarded as the marker gene for the cluster label. Do you think there are other reasons that the spatial samples would interfere with the genes included in the basis matrix? Thanks!