search

QuKunLab / SpatialBenchmarking

BSD 2-Clause "Simplified" License
107 stars 26 forks source link

PCC & SSIM in deconvoluting methods #3

Closed lzygenomics closed 2 years ago

lzygenomics commented 2 years ago

Hi~

Thanks for your nice work！ It's really fascinating~ I have a question about how to calculate the PCC and SSIM in deconvoluting methods~? The results of these methods seem just offer the proportions of cell types, so I wonder how to transfer proportion to genes expression ? : )

Best wishes!

wenruyustc commented 2 years ago

Hi. For cell type deconvolution, we constructed 32 simulation datasets with known cell type distributions as ground truth (uns['density'] is the cell type composition of each spot for simulation datasets) to evaluate the performance of these algorithms. So for each datasets, cell type composition of each spot by prediction can be used to calculate the correlation with cell type composition of each spot in ground truth.

lzygenomics commented 2 years ago

Hi. For cell type deconvolution, we constructed 32 simulation datasets with known cell type distributions as ground truth (uns['density'] is the cell type composition of each spot for simulation datasets) to evaluate the performance of these algorithms. So for each datasets, cell type composition of each spot by prediction can be used to calculate the correlation with cell type composition of each spot in ground truth.

Thanks for your quickly reply~ It's really helpful !