search

Danko-Lab / BayesPrism

A Fully Bayesian Inference of Tumor Microenvironment composition and gene expression
147 stars 46 forks source link

Using cross-species reference and combining references #27

Open alvaannett opened 1 year ago

alvaannett commented 1 year ago

Hi!

I have two questions about the sc reference:

  1. Is it ok to us data from different species, i.e. mouse sc data to deconvolute human bulk?
  2. Is it ok to mix sc data from different labs/publications etc in the reference or does this introduce batch effects that can not be accounted for?

Thank you!

tinyi commented 1 year ago

Thank you for your interest in our method.

To answer your questions: 1) I have not tested cross species deconvolution. I usually would not recommend this strategy, unless the marker genes of each cell type remain highly cell type-specific across species. 2) When the batch effect is small, it is ok to mix the reference. When the batch effect is huge, and one batch is more similar to bulk than the other, the batch that is more similar to bulk RNA-seq will get inflated cell type fractions. There is a workaround to this: if all cell types are present in every batch, one can treat each batch as a cell state, and BayesPrism will automatically determine the weights followed by summing it up.

Best,

Tinyi

On Tue, Feb 28, 2023 at 6:26 PM alvaannett @.***> wrote:

Hi!

I have two questions about the sc reference:

  1. Is it ok to us data from different species, i.e. mouse sc data to deconvolute human bulk?
  2. Is it ok to mix sc data from different labs/publications etc in the reference or does this introduce batch effects that can not be accounted for?

Thank you!

— Reply to this email directly, view it on GitHub https://github.com/Danko-Lab/BayesPrism/issues/27, or unsubscribe https://github.com/notifications/unsubscribe-auth/AB4NHS5TVHO5KI3CPPT4P4DWZ2CQZANCNFSM6AAAAAAVLLA3PU . You are receiving this because you are subscribed to this thread.Message ID: @.***>

alvaannett commented 1 year ago

Thank you, that's very helpful!