I am processing fungal LSU 454 data using Mothur 1.31. I used LR0R/LR3 to amplify ca.600bp of fungal LSU.
It really shocks me that the command pre.cluster (Environ Microbiol. 2010 July; 12(7): 1889–1898), which is crucial for reducing pyrosequencing errors, removes over 80% of good unique sequences from my samples. From the OTU list, most of my samples have 15-30 OTUs out of 5000ish reads per sample based on 0.03 cutoff.
I can accept if the fungal communities really are of low-diversity/extremely uneven. But I feel a bit insecure because the 'errors' that the pre.cluster detected may be due to the limit of 454 read-length.
I also looked up the pipline in RDP, seems like there's no precluster option.
What should I do now? Is there any alternative method I can use?
Would be interesting to know if they are detected as good fungal LSU sequences by Metaxa: http://microbiology.se/software/metaxa2/
What were the precluster settings you used? I know the mothur SOP recommends 1 base per 100, but that's really just a number empirically determined for 16S data.
When you say it removes 80% of the sequences (it doesn't actually remove then, it just merges the rare ones into the more abundant ones) how many OTUs/unique sequences does that leave you with?
- Badges
- Science method