Singletons -Amplicon sequencing

[Xterra]

We performed our first run using 14 MID -14 different specimens per sector with an 4-sector gasket. Under these conditions we are getting ~10,000 reads per specimen. Every seems to be fine with the reads however, we are getting plenty of singletons (40-60%). Thus, if we treat them as 'errors' we would end up with very few haplotypes. We know the complexity (# of haplotypes) is quite high in those samples though, and therefore having so few haplotypes does not make sense. If they are not errors, then their frequency is too low and that does not seem to be correct either. I was wondering what is the 'common' way to define a singleton when dealing with amplicon sequencing? In other words, are they commonly accepted as actual variants or discarded as errors?
X

[flxlex]

Are you doing similarity based clustering to get the haplotypes? Could these singletons be the result of sequencing errors? Perhaps using pyronoise program (http://www.nature.com/nmeth/journal/...meth.1361.html), also available as a part of QIIME (http://qiime.sourceforge.net), could help to reduce this noise?

[Xterra]

Pyronoise

We will give it a try.
Thanks