Hi
I need to combine two 454 datasets but one was obtained with FLX (400 nt) and the other with FLX+ (700 nt), both starting with the same forward primer. I know I have to rarefy to the lowest number of sequences to calculate indexes and compare between samples. But as far as I know, in Qiime for example, you can run a pipeline without screening all sequences to the same length (and not even aligning the sequences to a reference alignment before, as it is done in Mothur).
Is this then possible to compare diversity indexes and inspect beta diversity as well without having all sequences spanning the same length? meaning more variable regions for some sequences (some samples) and less for others?
Or as well as the sampling effort (number of sequences per sample), trimming to have the same sequence length and positions (same region in all sequences using E. coli gene as reference) is a "must do"?
Thanks for your feedback!
You can do whatever you want, question is does it make sense. Intuitively I would say it doesn't make sense because, depending on your OTU clustering algorithm, terminal gaps could be penalized. Thus sequences that otherwise would cluster to the same OTU (if looking at just one similar v region e.g. V3) would not cluster together if you don't trim. If I understand correctly this might inflate your OTU count as well as skew your beta diversity measurements.
This guy explains it well (for a physicist :- P)
http://drive5.com/usearch/manual/global_trimming.html
I also recommend reading his paper regarding uparse if haven't already.
http://www.nature.com/nmeth/journal/v10/n10/full/nmeth.2604.html
You absolutely should trim your sequences to the same length before clustering to OTUs. Both mothur and UPARSE consider terminal gaps to be differences (and QIIMEs main OTU picking strategies are based on UPARSE) so by allowing different lengths in you're artificially inflating the differences in your sequences.
Thank you a lot Ido and David for your answers!
I don´t have access to the Nature communication, but good to read the physicist explanation!
I thought it was only a matter of mistakes in calculation of diversity indexes due to more information for some organisms than for others, but that I would gain in accuracy of taxonomic identification of OTUs, but it seems the problem is deeper and the very OTUs building can be compromised. I knew I had to trim all sequences to the same length using Mothur due to the alignment strategy it uses but I was not aware about the problem with UPARSE in Qiime.
- Badges
- Science topic
- Similar topics
- Bioinformatics
- Bioinformatics and Computational Biology