5 replies to this topic

[littleaxt]

Hi

I calculated my normalization factor (NF) in excel and just wondered why it was different from the one in geNorm. It seems geNorm normalized the expression data to the highest expression value in each gene which seems rather strange to me as it then compares the genes within the sample with each other, allthough they are all differently normalized.
Isn't it better to normalize within the sample, than within the gene?

[gfischer]

geNorm uses its own statistical methods for determining the best HKG by comparing the candidates to each other.  I really don't know the statistics behind it, but here's two papers validating the method, along with another program, NormFinder

Attached Files

•  Ohashi_BBRC_2008.pdf   451.09K   133 downloads
•  Ohashi_BBRC_2008a.pdf   196.63K   172 downloads

[littleaxt]

gfischer, on Jun 15 2009, 04:17 PM, said:

geNorm uses its own statistical methods for determining the best HKG by comparing the candidates to each other.  I really don't know the statistics behind it, but here's two papers validating the method, along with another program, NormFinder

Thanx!
I also tested BestKeeper, NormFinder and geNorm and they mainly produced the same results.
Vandesompele et al. use the geometric mean but if you look at the example data that they provide they had normalized the data.
That's basicly fine, but if they compare each gene in a sample to another gene in the same sample and these genes are differently normalized it seems inaccurate to me. Or am I wrong?
Maybe I should ask themselves.

Cheers!

[gfischer]

Like I said, I don't claim to understand the math behind systems like GeNorm, but from what I understand, they work on the principle of working within the sample set and comparing each gene to all the other genes.  Soehow, it can calculate the relative variance of each gene compared to the others and determine the most stable gene.  You'd probably have to talk to a biostatistician for an explanation of why it works.

[Rsm]

They also have a discussion group on Yahoo here, where you can get very good feedback...
I did not use their program, because they assume that HKG have similar expression between samples, just the distribution is different (one high, one low and reverse in your other sample). In my case I had down regulation of almost all HKG in one sample (and I've tried more than 20), of course using the same amount of RNA. In this case you can not use those programs....
Cheers,
Minna

[littleaxt]

Hi!

Thanx for your replies. I'll ask in the yahoo-forum. If I get an answer it can post it here.

All the best!

Jan