DEA best practices: raw vs. normalized data
Hi all,
I am searching for best practices in DEA using scanpy.
I can't find it back but I remember reading somewhere that DEA should be done on raw expression data rather than normalized data. However in scanpy tutorials they seem to perform it on log normalized data. Moreover, scanpy.tl.rank_genes_groups throws a Warning if it does not recognize log transformed data (which I think is because the Fold Change is computed using an exponential to reverse the log, so if it has raw data it applies an exp to it which is not logical).
Thank you for any precious insight,
0 answers
No answers yet.
Log in to answer this question.
More posts like this
-
Appropriate normalization of skewed data for "Translating multiomics single-cell data" with MOSCOT
written by npont •
-
Tutorial: Single-cell RNA-seq: Annotation: Celltype auto annotation with MetaTiME
written by Julia Ma •
-
Tutorial: Bulk RNA-seq: Different Expression Analysis
written by Julia Ma •
-
Tutorial: How to preprocess and visualize beautifully scRNA-seq with omicverse?
written by Starlitnightly •
-
Tutorial: Unsupervised clustering on gene expression data
written by Hamid Ghaedi
-
How to filter normalized RNA-Seq transcripts for the purpose of predicting a continuous phenotype
written by arf1389 •
-
In Which Type Of Data Does Combat Best Perform?
written by fbrundu
-
Tutorial: Transforming And Manipulating Color Space Reads
written by Istvan Albert
Depends on the test. If you use tests that use data "as-is", for example t-test, Wilcoxon, limma-trend, etc, then normalization must be done before. In almost all cases you want to log-transform. Other tests such as edgeR use the normalization as part of their model with offsets.