cell state label is mandatory ?

[martina811]

Hi, I would like to use BayesPrism on rnaseq data, is it mandatory to have also a file which specify the cell state label of my single cell data? Or can it work with just the cell type label ?

many thanks

[tinyi]

Yes. To only use cell type, simply specify cell.state.labels = the same input you provided for cell.type.labels. Theoretically speaking, the definition of cell states were developed to account for the uncertainty in the reference, particularly for cell type whose distribution is heterogeneous and lies on a continuum, and hence was not developed for accurate quantification. The reason for this is that suppose cell state B is a linear combination of cell state A and C, even though they are presented in the same amount in the mixture, the inference can be unstable (prone to the data quality of the cells that constitute the cell state), and the sparse prior may also penalize cellstate B to close to 0.

…

On Tue, Nov 14, 2023 at 12:27 AM martina811 ***@***.***> wrote: Hi, I would like to use BayesPrism on rnaseq data, is it mandatory to have also a file which specify the cell state label of my single cell data? Or can it work with just the cell type label ? many thanks — Reply to this email directly, view it on GitHub <https://nam12.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2FDanko-Lab%2FBayesPrism%2Fissues%2F66&data=05%7C01%7Ctc532%40g.cornell.edu%7Ca95a1e0d20fa404fad4408dbe465736f%7C5d7e43661b9b45cf8e79b14b27df46e1%7C0%7C0%7C638354896609810557%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=w5aT2NpX38QShaqLQvzfHDa6kep89FJ3nBCimm3gy5Q%3D&reserved=0>, or unsubscribe <https://nam12.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FAB4NHS5EWSQM5OGDBLT5ABDYEJC7TAVCNFSM6AAAAAA7JNOUQCVHI2DSMVQWIX3LMV43ASLTON2WKOZRHE4TCMBRGAYDAOA&data=05%7C01%7Ctc532%40g.cornell.edu%7Ca95a1e0d20fa404fad4408dbe465736f%7C5d7e43661b9b45cf8e79b14b27df46e1%7C0%7C0%7C638354896609810557%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=5daiPWoAt0eaPZ9pzoWnrJkh0vptcIsZ5Awy0i06%2BKA%3D&reserved=0> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[yunbokai]

Hi @tinyi , I want to estimate and compare the fraction of Tcell_sub1, Tcell_sub2 and Tcell_sub3 in different RNA-seq data. Is it reasonable to set Endothelial, Fibroblats, Epi, Myeloids, Tcell_sub1, Tcell_sub2 and Tcell_sub3 as cell type label and set Endothelial_sub1, Endothelial_sub2, Fibroblats_sub1, Fibroblats_sub2, Epi_sub1, Epi_sub2, Myeloids_sub1, Myeloids_sub2, Tcell_sub1, Tcell_sub2 and Tcell_sub3 as cell state label? Or do you have any other suggestion? Thanks.