Integration with other scRNA datasets

Question

Integration with other scRNA datasets

jindalk opened this issue 4 years ago · comments

Hi, thanks a lot for making your data available on github. I'm planning to integrate the in vitro data with existing 10x scRNA datasets. Do you have any recommendations on how to go about that, given that the counts provided are already normalized?

Thanks

Caleb Weinreb · Answer 1 · Thu Oct 15 2020 01:15:39 GMT+0800 (China Standard Time)

It might be best to use a published approach for batch correction / combining scSeq datasets. Typically when I combine datasets, I make sure they are total counts normalized so that cells in both datasets have the same total counts, and then we do "PCA-batch correction" which is described in the supp methods of "Lineage tracing on transcriptional landscapes links state to fate during differentiation"

…

On Wed, Oct 14, 2020 at 11:56 AM jindalk ***@***.***> wrote: Hi, thanks a lot for making your data available on github. I'm planning to integrate the in vitro data with existing 10x scRNA datasets. Do you have any recommendations on how to go about that, given that the counts provided are already normalized? Thanks — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#7>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABO45MR4WJKZI7ZTWVLGLPLSKXC2DANCNFSM4SQZQ4FA> .

velten-lab · Answer 2 · Fri Feb 05 2021 19:21:49 GMT+0800 (China Standard Time)

Along the same lines, I like to work with integration methods that start with raw count tables. If I understand you right, the files you provide contain library size normalized values. Could you additionally provide the raw count tables?
Thanks and greetings, Lars

Caleb Weinreb · Answer 3 · Sat Feb 06 2021 08:53:17 GMT+0800 (China Standard Time)

The raw counts are not available but you can recover them from the normalized counts. During normalization each cell was uniformly scaled by common factor. You can recover this factor by taking the reciprocal of the smallest non-zero gene-expression value for that cell. You should then be able to rescale each cell such that all the gene expression values are integers.

…

On Fri, Feb 5, 2021 at 6:22 AM velten-lab ***@***.***> wrote: Along the same lines, I like to work with integration methods that start with raw count tables. If I understand you right, the files you provide contain library size normalized values. Could you additionally provide the raw count tables? Thanks and greetings, Lars — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#7 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABO45MXUE6CSAYYSXFJUJS3S5PIFZANCNFSM4SQZQ4FA> .