search

neurorestore / Augur

Cell type prioritization in single-cell data
MIT License
94 stars 10 forks source link

Question about comparing AUCs within the same Augur object #31

Open alicedini opened 7 months ago

alicedini commented 7 months ago

Hello, thank you for the amazing tool! This might be a silly question for you, but would it make sense to compare the AUC scores obtained for the different cell types within the same Augur object? For instance, could I use the DeLong test? And should I use Augur in the default mode, or with the permute one?

Thanks in advance for your help, Alice

skinnider commented 7 months ago

Hi @alicedini, we do have some functionality for comparing AUCs in Augur, see the protocol here https://www.nature.com/articles/s41596-021-00561-x especially case study 4 and box 3. Does this help?

DelongZHOU commented 3 months ago

Hello, Following up on this question. Reading the manuscript, it seems to me that the permute option seems to focus on comparison of the same cell type accrose sex or treatment. I believe both the original post and me want to compare between celltypes within the same sex, treatment etc. At least for me I have two related questions. Suppose for one augur run, I have AUC=0.60 for cell type A and AUC=0.52 for cell type B. Is each of these AUC>0.5 statistically significant? Is AUC_A > AUC_B statistically significant? Can I use the iteration results in the $result dataframe as random sampling replicates? Thank you!