Is it appropriate to apply quantile normalization on RNA-Seq before computing the Pearson correlation?
Is it appropriate to apply quantile normalization on RNA-Seq before computing the Pearson correlation? Specifically, RNA-Seq data displays a dense tail distribution, and Person's correlation (cor function in R) assumes a normal distribution. I tried this in R using the normalize.quantiles function discussed in the lecture, and my results make a lot of sense, but I want to make sure that this is an appropriate transformation before accepting these results.
Thank you for your help!
0 answers
No answers yet.
Log in to answer this question.
More posts like this
-
SCTransform vs LogNormalize for preprocessing scRNA-seq data before cMonkey biclustering
written by arjunp22 •
-
What is the right choice of normalization and preprocessing before WGCNA using RNAseq data in a coh…
written by sukeshinik5 •
-
normalization RNASeq Data for QTL analysis
written by Linda •
-
Tumour heterogeneity in scRNA-seq - cell-to-cell correlation
written by dccroucher •
-
Which one to use for ChipSeq replicate comparison: Spearman or Pearson
written by rohitsatyam102
-
Correlation heatmaps: use correlation as clustering distance or not?
written by CuriousGuy •
-
How to avoid duplicate edges between same node pair in a co-expression network
written by aishu.jp •
-
How to manage self loops in gene coexpression network constructed from RNA-seq data
written by aishu.jp •
-
Tissue Specificity R-package
written by BioinfGuru
-
Correlated genes in two different datasets
written by salvatore.raieli2
Of course you will get normal distributions out of
normalize.quantilesbut I wouldn't count on it. The main problem is the low counts, you will get random effects amplified. I would try spearman's correlation instead but make sure you normalize for library size first.Yes. My pipeline is rawcounts-->DESeq2-->normalized.quantiles Or do you suggest normalizing the rawcounts using a different method? Is there an R function to do the library normalization?
DESeq2 normalization is great. Maybe try spearman's on the normalized count table. Make sure the
rdistribution looks good. You can use thenormalize.quantilesbut just make sure you're not getting artifacts from 0 counts.Yes. I am removing rowMeans<10.