Question about using BayesPrism with TCGA bulk data and 5'/3' single-cell references

[Rapamycin-s]

Hi Tinyi,

Thank you for developing and maintaining BayesPrism — it's a powerful and well-documented tool.

I’m currently trying to use BayesPrism to deconvolute TCGA bulk RNA-seq data using single-cell RNA-seq data as a reference. I have a question regarding the impact of using 5’ versus 3’ single-cell RNA-seq data as the reference input.

Does the use of 5’ or 3’ single-cell RNA-seq (e.g., from 10x Genomics) significantly affect the deconvolution results when applying BayesPrism to bulk RNA-seq data (like TCGA)? Is one more suitable or recommended than the other?

I understand that BayesPrism models gene expression distributions, so I was wondering whether the capture bias introduced by 5'/3' protocols would skew the reference profiles or have any downstream impact.

Any guidance on best practices or relevant references would be greatly appreciated.

Thanks in advance for your time!

Best regards,
Jayden

[tinyi]

Dear Jayden, Thank you for your interest in our methods. This is a good question, and I do not have a straightforward answer. I have mostly used 3'end scRNA-seq since they present the major type of scRNA-seq data. You can try both types of scRNA-seq dataset and see how they differ. You may also concatenate the scRNA-seq reference and see if BayesPrism gives higher weight for a particular sc platform - especially for the same cell type. Best, Tinyi

…

On Wed, Apr 30, 2025 at 8:48 PM Rapamycin-s ***@***.***> wrote: *Rapamycin-s* created an issue (Danko-Lab/BayesPrism#116) <#116> Hi Tinyi, Thank you for developing and maintaining BayesPrism — it's a powerful and well-documented tool. I’m currently trying to use BayesPrism to deconvolute TCGA bulk RNA-seq data using single-cell RNA-seq data as a reference. I have a question regarding the impact of using 5’ versus 3’ single-cell RNA-seq data as the reference input. Does the use of 5’ or 3’ single-cell RNA-seq (e.g., from 10x Genomics) significantly affect the deconvolution results when applying BayesPrism to bulk RNA-seq data (like TCGA)? Is one more suitable or recommended than the other? I understand that BayesPrism models gene expression distributions, so I was wondering whether the capture bias introduced by 5'/3' protocols would skew the reference profiles or have any downstream impact. Any guidance on best practices or relevant references would be greatly appreciated. Thanks in advance for your time! Best regards, Jayden — Reply to this email directly, view it on GitHub <#116>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AB4NHSZ2VN3DFNAB26WM4E324FVO7AVCNFSM6AAAAAB4G566U2VHI2DSMVQWIX3LMV43ASLTON2WKOZTGAZTEOBYGUYDEMY> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[Rapamycin-s]

Dear Tinyi,

Thank you so much for your prompt and thoughtful response.

I appreciate your suggestion to compare both 5’ and 3’ single-cell RNA-seq datasets, and to potentially concatenate them to assess platform-specific weighting within BayesPrism. That’s a very helpful and practical approach — I’ll certainly give it a try in my analysis.

Thank you again for your time and for developing such a valuable tool. I’ve found BayesPrism to be incredibly insightful for my work.

Best regards,
Jayden

[tinyi]

You are welcome, Jayden!