Merging (or not) ITS2 amplicons with variable length

[colinbrislawn (Colin J Brislawn)]

Hello @rahel_park

@rahel_park

In the same run there were indeed some mock samples as well. I added them now to the previously tested samples. The mock community was made by mixing together in equal cell number 13 different yeast and extracting the DNA with the same kit as used for the rest of the samples.

Well done! I'm so glad you included that control. This let's us answer LOTS of important questions.

@rahel_park

The amplicon length for the mock community is ranging between 150 to 369 bp

This is one of the standing challenges with ITS reads. The variable length region can confuses pipelines that expected a constant length region like v4 16S.

@rahel_park

should normally be well covered by MiSeq 2x300 reads

This means that some reads can overlap by (300x2 - 369) = 231 overlap or totally overlap with overhang on both sides. This overhang with +100% overlapping reads can also confuse some joing steps. I'm not sure how well dada2 works with fully overlapping reads.

@rahel_park

When the ITSX is used and --p-trunc-q 20 is not used in denoising, then the longest amplicon does appear in the table, but 10 times less abundantly compared to forward only.

I wonder if this is due to the trunc option clipping the reads, so that they don't join and don't appear in the data. But even if you don't truncate your reads, they might still not join!! As I discovered here, the qiime2 dada2 plugin expects zero missmatches in the reads when joining. This would means a drop in quality could prevent joining, with or without the trunc option.

Maybe we should change the dada2 default to allow more than one missmatch so reads can actually join...

---

@rahel_park

I am considering to do all the analyses with forward only and then as well ITSX and without the extra --p-trunc-q 20 in denoising - and compare how much the end results differ.

I really like this idea. You can use the mock community to choose settings that fully capture the diversity in your samples.

Let me know what you find!

Colin