Ngs Data Normalization

Question

[jack]

Hi all,

I have RNA seq miRNA data set(NGS). it's it .txt format. I want to check that, wthether this data are normolazed or not. could sombody let me know how can I check it in R?

1    3    0    2    0    1    0
0    0    0    0    0    0    0
0    0    1    0    0    0    0
0    4    0    0    0    0    0
579    1494    401    321    65    190    39
0    0    0    0    0    0    0
0    0    0    0    0    0    0
0    0    0    0    0    0    0
0    0    0    1    0    0    0
0    0    0    0    0    0    0
0    1    0    0    0    0    0
1    1    1    1    0    0    1
41    163    42    113    25    57    14
1    1    1    0    0    0    0
44    14    25    10    6    12    5
0    0    0    0    0    0    0
0    0    0    0    0    0    0
0    0    0    0    0    0    0
0    1    0    0    0    0    0
1    0    1    0    0    0    0
0    0    0    0    0    0    0
12    10    3    2    2    5    0

[Devon Ryan]

Could you show a few lines of the text file so we know what format the data is in? Your question is impossible to answer otherwise.

Update: Thanks, that's exactly what's needed!

[Rm]

Looks like data table parsed from the mirdeep2 exp files; Use EdgeR or DESEQ to normalize and do DE analysis.

check this article on NGS RNASeq normalization: http://bib.oxfordjournals.org/content/14/6/671.full

[JC]

You can try many scaling/normalization algorithms with Normalizer: http://db.systemsbiology.net/gestalt/normalizer/

Answer 1

[Devon Ryan]

Those appear to be raw counts per miRNA. You'll want to use DESeq/edgeR/limma (see the voom() function) or one of the other standard tools to continue with the analysis.

[jack]

I want to do some clustering with my data, but before it, i want to do quantile normalization. voom() seems doesn't support it.

[Devon Ryan]

There's no reason for voom() to do that, it just transforms things for use in limma. If you want to do quantile normalization, just do it after importing counts with voom() (since doing quantile normalization on raw counts wouldn't make much sense).

[jack]

I don't want to use limama, I have my own algorithm. but I don't know much about NGS data, I have the above raw data and I want to normalize it and give it to my algorithm. for micoarray data, I already used quantile normalized data, but for NGS, I'm not sure what should I use . do you have any recommendation ?

[Devon Ryan]

Then just perform quantile normalization on the transformed counts.

[jack]

which kind of transformation i should do before quantile normalization? is the quantile normalization is the best normalization method for NGS data ? my algorithms basically will do first clustering then dimentionality reduction and finally visulazation. I don't want to miss biological patterns due to choosing wrong normalization pattern .

[Devon Ryan]

Your algorithm sounds like combat. Regarding transformation, either voom or vst (from DESeq) would work. I find the concept of a single "best" normalization method for all use cases rather simplistic.

[jack]

thanks, so what is exactly 'single "best" normalization method''?

[Devon Ryan]

There isn't one