search

navinlabcode / copykat

Other
214 stars 55 forks source link

using normalized data as input #33

Closed zhangguy closed 2 years ago

zhangguy commented 3 years ago

Hi Ruli,

Thanks so much for developing this method!

I was wondering, if I don't have access to the raw UMI counts, but just have lognorm counts, is there a way I can still use this method? I understand that copyKat does normalization on the UMI using the Freeman-Tukey transformation, and theoretically it wouldn't make sense if I use log(norm UMI) as input. It would be highly appreciated if you could suggest a way for lognorm data or just normalized data. Thank you so much.

gaobio commented 3 years ago

Hi Ruli,

Thanks so much for developing this method!

I was wondering, if I don't have access to the raw UMI counts, but just have lognorm counts, is there a way I can still use this method? I understand that copyKat does normalization on the UMI using the Freeman-Tukey transformation, and theoretically it wouldn't make sense if I use log(norm UMI) as input. It would be highly appreciated if you could suggest a way for lognorm data or just normalized data. Thank you so much.

Yes, lognorm might work. You can have a test. We used log(TPM+1) for Smartseq2 data, which worked well.