cells clustering by sample

[michael-vermeulen]

My dataset contains one cell type from 10 samples. Each sample was sequenced independently and therefore is its own batch as well. When I cluster using desc each sample clusters purely by sample (with minimal overlap)

I have tried varying the learning rate with limited success. Are there other variables to adjust the model to encourage more rigorous batch correction?

[eleozzr]

Can you try scaling data within each batch and then feed the scaled data into desc?