several samples were analysed by T-RFLP; the resulting spectra were clustered using Pearson correlation as a distance measure. these clusters correspond well with the theoretical model.
...almost :)
"almost" because there is an interval in T-RFLP spectra (about 10% of all peaks) which approximately corresponds to a certain bacterial phylotype (according to the clone library), but the peaks inside this interrval are problematic: they are quite intensive, but not very stable, changing even between replicates. the clustering only becomes good enough when these peaks are discarded.
Some tells me that throwing them away is a bad idea because T-RF's are dependent variables and cannot be treated as independent ones for community comparison. Is it so, in this case?
How else can one reduce the effects of dominating OTUs on ordination, and identify possible indicator T-RFs for certain effects within the dataset?
One could argue this is truly not a bioinformatics question, as your question is a technical one: how to visually call multiple Sanger sequencing peaks for T-RFLPs. This is part of the reason why I haven't done T-RFLPs in 6 of so years, because it's a troublesome technique that leaves a lot of choice up to the researcher (i.e. is this sequence peak the combination of 3 individuals or 300 individuals?). If you have dominant OTUs they will swamp out the rest of the peaks, which is difficult to interpret. I think an amplicon sequencing experiment and/or RT-PCR should always go hand-in-hand with T-RFLPs to validate them.
I don't have much advice for you as its difficult to call T-RFLPs, but there are a few packages that can help you sort out the signal from the noise. I would recommend the TRAMPR package in R. I've never used the T-RFLP STATS package, but it uses both Perl and R. This T-RFLP analysis cookbook has some good scripts that are a bit dated.
Good luck with your analysis! Feel free to contact me if you have any questions!
