search

txWang / BERMUDA

BERMUDA (Batch Effect ReMoval Using Deep Autoencoders) is a novel transfer-learning-based method for batch correction in scRNA-seq data.
MIT License
27 stars 7 forks source link

confusion about the Supplementary information of BERMUDA #9

Open yuxiaokang-source opened 3 years ago

yuxiaokang-source commented 3 years ago

In the supplementary information(https://static-content.springer.com/esm/art%3A10.1186%2Fs13059-019-1764-6/MediaObjects/13059_2019_1764_MOESM1_ESM.pdf) of the paper(BERMUDA: a novel deep transfer learning method for single-cell RNA sequencing batch correction reveals hidden high-resolution cellular subtypes) image

why the experiment removal 1 have 3 figures(c2-4,I think it should be d2-4), what is the difference between c2,c3,c4(d2-4),the same question happend on experment removal2 ? in my opinion, It is enough for one experiment corresponds to one figure? Could you give some explain about this?

txWang commented 3 years ago

Thanks for the comment. Yes, it should be d2-4. Sorry for the confusion. In the case of Experiment removal1 and Experiment removal2, the performance was evaluated with 3 scores. Since it is hard to clearly visualize 3-D scatter plots here, we plotted 2-D scatter plots for each pair of scores. Note that the axes in d2-4 are different. For Experiment all, the performance was only evaluated with 2 scores. There is no entropy_score as there is no batch-specific cell type in Experiment all.