Best Practice for Analyzing Multiple Single-cell Sequencing Samples with CellDancer

[weir12]

Hello CellDancer Developers,

I'm working with multiple single-cell sequencing samples and am seeking advice on the best approach to analyze RNA velocity using CellDancer. Should I merge the samples and run CellDancer once on the combined dataset, or is it advisable to run CellDancer separately for each batch?

Additionally, I'm concerned about potential batch effects, similar to those encountered with expression matrices. Can the RNA velocity results from different batches be directly compared, or are there considerations for batch effects that need to be addressed?

Thank you for your guidance on this matter.

[Abclisy]

Hello,

Thank you for reaching out. When running the prediction, we assume the batch effect is already corrected. From my own experience, you can observe both ’the embedding space‘ and ‘the unspliced-spliced phase portrait of some genes’ to check the batch effect. For removing batch effect, there are many methods https://www.nature.com/articles/s41592-021-01336-8
Practically, you can correct the batch effect and verify it in the embedding space. Then decide whether to run the velocity prediction for each batch based on your observation of the phase portrait.

Best,
Shengyu