Open TomKellyGenetics opened 6 years ago
You should try not scaling the data when calling PCA. sklearn implementation only centers the data.
I've updated this to be consistent with the Python version. Unfortunately, the issue persists and the results are still considerably different.
Doublets are only identified with relaxed p-value thresholds (not the default of
p <=0.01
). While these are significantly enriched for doublets identified by the Python implementation (by Fisher's Exact Test), many different cells are identified in the same datasets.