Use latent embeddings for clustering

[aelhossiny]

Thanks for such an incredible method. I have a question about the clustering, are there specific reason for using the reconstructed gene expression matrix but not the learned latent embedding for the clustering?

[longyahui]

Hi,

Thank you for raising this critical issue. Simply, the reconstructed expression is more informative than the latent representation. For more explanations, please kindly follow the response 1.5 in the rebuttal file (https://static-content.springer.com/esm/art%3A10.1038%2Fs41467-023-36796-3/MediaObjects/41467_2023_36796_MOESM3_ESM.pdf). Hope it helps. 

Best regards,
Yahui