Question about calculate the immune cell fraction in bulk sequencing of tumor sample

[threshold233]

Dear developers:

First of all, thank you for developing this great algorithms, it really helped my research.
I try to find the specific immune cell fraction in a bulk dataset containing multiple cancer.
But in this project, the reference scRNA dataset just contain 3 cancer type, which means only 3 types of cancer cell can be included as the reference for buliding the Prism object. But the bulk dataset containing 10 cancers.
So i wonder can i just use the non-tumor cells as the reference in buliding the Prism object and what problem may this method cause in using BayesPrism if i do not include the tumor cell.

Thanks for your help!

[tinyi]

Dear user, Thank you for your interest in our methods. It is always recommended to use the most comprehensive representation of cell types to build the scRNA-seq reference, as the statistical model assumes that all cell types are found in the reference and their fractions sum to one. If only using non-tumor cells, the BayesPrism will distribute reads from tumor cell to the cell type that are most similar to tumor cell in the reference, and hence inflate its fraction. I was not quite sure what is exactly meant when you say 3 cancer type. Typically, it is recommended to deconvolve one cancer type, e.g. GBM or lung cancer, at a time, using the scRNA-seq from the matched cancer type to build the reference. Best, Tinyi

…

On Mon, Aug 12, 2024 at 5:11 AM threshold233 ***@***.***> wrote: Dear developers: First of all, thank you for developing this great algorithms, it really helped my research. I try to find the specific immune cell fraction in a bulk dataset containing multiple cancer. But in this project, the reference scRNA dataset just contain 3 cancer type, which means only 3 types of cancer cell can be included as the reference for buliding the Prism object. But the bulk dataset containing 10 cancers. So i wonder can i just use the non-tumor cells as the reference in buliding the Prism object and what problem may this method cause in using BayesPrism if i do not include the tumor cell. Thanks for your help! — Reply to this email directly, view it on GitHub <#95>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AB4NHS26LPWPNFHYYDP4B4LZRB34HAVCNFSM6AAAAABMLX7XRWVHI2DSMVQWIX3LMV43ASLTON2WKOZSGQ3DANBUGE4DSNI> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[threshold233]

Thank you for your explain, I know what i should for my assay.