Entering edit mode
2.1 years ago
fifty_fifty
▴
60
I used two different methods for dimensionality reduction before Seurat clustering. Here are UMAP visualization plots for each of my pipelines. One is denser than another, but still, both pipelines clustered cells of the same type together. What is the best way to compare the results in this case?
Isn't one of the main points of this sort of analysis to find the most specific markers for each cell type? If you find markers for each data set, and compare them, are the lists the same for each cluster? Does one set have better characteristics of specificity than the other?
usually, yes. But in this case, I already have the labels. So, I just need to assess which dimensionality reduction method worked better.
Labels and marker genes are not the same thing. Nevertheless, do you have a definition of what it means to "work better" or are you looking for one? There's a recent publication on Benchmarking Clustering that discusses some of the issues.