I am analysing data from 30 DGGE images using GelCompar. The banding patterns are quite complex so I have been doing a curve-based analysis. The methodology in published literature seems to be to use Pearson correlation and then UPGMA for clustering. I have also been debating whether to use cosine similarity over the Pearson correlation, but I am struggling to find clear explanations of the differences between the two and what they do to make this decision so any input on this would be appreciated. My data is sheep milk samples collected from each half of the mammary gland over eight consecutive weeks and then the samples for each ewe have been run on individual DGGE gels.
Secondly, I am trying to find a way to export the values GelCompar uses to produce the densitometric curves for the curved-based analysis to use in other programmes. Has anyone done this before? How?
Hello Emma,
I'm sorry that I can't help you with your question about the statistical differences between Pearson correlation and cosine similarity. In our lab we're using the Pearson correlation to compare different PCR-SSCP band profiles of bacterial or fungal communities in dust, milk, feces and more. It works quite well.
For the export of the densitometric values from GelCompar II start a comparison of your DGGE band profiles, click on 'show curves' and export the densitometric values by opening the 'File'-menu at the top of the comparison window and then click on 'export densitometric curves' in the drop-down menu. This function is only active if the 'show curves' button is pressed before. You'll reveive a txt-file where you can copy the values into MS Excel.
Kind regards,
Tobias
Dear Emma,
you should consider googling: Numerical Ecology with R pdf
there is a free copy of this book on www, with plenty of explanations on when, how and why use different metrics.
Best,
Blaz
Thanks for the information Blaz, I'll have a look at the suggested link.
- Badges
- Science method