BayesPrism results from your GBM tutorial #125
Description
Hi, Thank you for this awesome tool first of all ! I have a few questions, would be amazing if you could help me.
1 - I ran BayesPrism using the same data from the GBM tutorial (Tutorial: bulk RNA-seq deconvolution using BayesPrism), but I don’t have a way to evaluate my results. Could you please share the fraction estimates (and, if possible, the per-cell-type estimated expression outputs) from the tutorial? That would be extremely helpful. Thanks !
2 - If you have any suggestions on how best to evaluate or validate the estimated cell-type fractions from BayesPrism (for example, recommended correlation metrics, consistency checks, or sanity diagnostics you typically use), I would really appreciate your advice. As a reference, I am currently using matched single-cell and bulk datasets (same patients in both datasets). However, such datasets are relatively rare, and even when derived from the same dissociated tissue, the cell-type compositions are not perfectly identical between the single-cell and bulk samples.
3 - Based on the matched dataset I am currently using, I observe that BayesPrism tends to underestimate some cell types (notably T cells) and overestimate others (such as endothelial cells). I am trying to understand the drivers of this behavior and explore ways to obtain more accurate fraction estimates; any insight you could share would be greatly appreciated. Thank you !!