omnideconv / deconvBench

Comparison of 2nd generation deconvolution methods implemented in omnideconv
2 stars 0 forks source link

Benchmarking fine-grained deconvolution #11

Closed FFinotello closed 9 months ago

FFinotello commented 2 years ago

Hello crew! @mlist and I were discussing the fact that 2nd-generation deconvolution could in principle unlock the quantification of closely related human immune cell types (e.g. T cell subtypes).

To test this, we could set up a simplified simulation-based benchmarking using the same single-cell data for signature building and pseudobulk data simulation. Maybe even without mRNA bias. In this simplified scenario, we could consider some fine-grained classification of T cell subtypes (e.g. Th1/Th2, or naive/activated/exhausted) to assess the methods. Maybe @kathireinisch could run this evaluation for her project? With the support of @alex-d13 with what regards the new simulator, of course.

I do not have any dataset in mind for Th1/Th2, but this is a nice one with annotated T cell subtypes including exhausted T cells: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE179994

Any other ideas?

Cheers, Francesca