Chimera check 250 bp sequences?


Whats the current thinking on using chimera checking for 16S rRNA iTAG sequences that are only 250bp long? From what I understand, chimera checking algorithms would have trouble accurately calling such sort sequences.


We do it all the time with MiSeq data.

So we did a test of this by taking a set of full length 16S sequences and then cutting them down to the size of our EMB Project primer set in silico, then comparing the chimera detection rates in mothur of the two data sets.

We found 2 actual chimeras in the full length set, but 17 (15 non chimeras) in the shorter (250bp) set of the exact same sequences.

This suggests that you could be throwing out a lot of non-chimeric sequences if the rates of detection on the shorter fragments are that far off from full length.


If you look back at the original UCHIME paper, they tested a range of fragment lengths and when you balance the sensitivity and specificity of chimera detection, the default parameters work pretty well.