UMI per million and RPM per million

Question

[wt215]

Hi,

I have two scRNAseq datasets. One is UMI based and another is non-UMI based.

The raw datasets are of either UMI per million or RPM per million format.

Is it possible to transform them back to the original count matrix? (For UMI data, the original count matrix should be consist of integers)

Or can I just round both datasets and put them into BASiCS or SCDE for normalization? As far as I know, those packages need original datasets as input.

Thank you very much!

Wenhao