search

jingshuw / SAVERX

R package for transfer learning of single-cell RNA-seq denoising
29 stars 9 forks source link

Input library-size-normalized expression #18

Open Winnie09 opened 4 years ago

Winnie09 commented 4 years ago

Hi,

Thank you for developing SAVERX. I am trying to impute multiple-sample single-cell UMI data using SAVERX. In order to make the gene expression comparable across samples, I have already normalized the UMI by library size. Now I want to impute the cells of each individual sample separately using these library-size-normalized expression, but SAVERX paper states that the input is UMI count. May I know whether we could input library-size-normalized expression to SAVERX, and how that would affect the imputation performance compared to using original UMI count?

Thank you! Winnie

jingshuw commented 4 years ago

Hi Winnie,

The input of SAVERX is UMI counts in order to do shrinkage. SAVER-X will do normalization of your data by library size, however we need the raw counts to correctly compute the loss function of autoencoder and the likelihood for shrinkage. If you input the normalized data, you will likely not get a correct answer. Based on my understanding, your ideal normalization is also by cell size, so SAVER-X will do the same normalization as you wanted.