[GenoMax]

Tags 1,2,3,4,5, and 7 worked
Tags 6,8,9,10,11, and 12 failed

I can't really see any pattern...

Were your tags radically different than the sample pool that ran on this flowcell (since you spiked these samples in)?

[Jessica_L]

I don't really see a pattern in the tag/plate failures, either. But if I understand correctly, the 6, 8, 9, 10 tags that failed the first time also failed the second? And Tag 7, which worked the first time, worked this time, too?

The only thing that makes sense to me here is that you're having an issue with the structure of your custom primers in the first PCR (the ones that are adaptor + 5bp tag + primer). I assume you've double checked all the sequences and didn't reverse complement something you weren't supposed to, otherwise you probably wouldn't have a 50% success rate. But have you checked the primers for things like dimers and hairpin loops? Given the fact that you have a low number of reads mapping to the failed tags (~50 reads per sample, you said in your first post) that could be random errors causing misidentification or it may be the result of a strong primer dimerization with only a few oligos available for amplification.

[vanessamata]

Were your tags radically different than the sample pool that ran on this flowcell (since you spiked these samples in)?

The samples that ran on this flowcell were of different sources, but yes, they were different from these ones with tags...

[vanessamata]

I don't really see a pattern in the tag/plate failures, either. But if I understand correctly, the 6, 8, 9, 10 tags that failed the first time also failed the second? And Tag 7, which worked the first time, worked this time, too?

The only thing that makes sense to me here is that you're having an issue with the structure of your custom primers in the first PCR (the ones that are adaptor + 5bp tag + primer). I assume you've double checked all the sequences and didn't reverse complement something you weren't supposed to, otherwise you probably wouldn't have a 50% success rate. But have you checked the primers for things like dimers and hairpin loops? Given the fact that you have a low number of reads mapping to the failed tags (~50 reads per sample, you said in your first post) that could be random errors causing misidentification or it may be the result of a strong primer dimerization with only a few oligos available for amplification.

Yes. 6,8,9, and 10 failed on both runs. And 7 worked on both runs.

Yes, the only thing that makes sense to me is that for some (molecular) reason the structure of these primers/tags are not allowing the sequencing to occur. I tested everything on agarose gel. I have no primer dimer and the amplifications are all of the expected size (and the same) with all primers/tags. They all incorporate the nextera index as well.

[Jessica_L]

I'll admit, this one has me stumped. If the library construction appears to have been successful, these should be sequencing. A few last questions:

Have you qPCR'd the individual libraries with the 5-bp tags? i.e. a separate set of reactions for each of tags 6, 8, 9, 10 with no pooling? It sounded like when you did the qPCR initially that it was on multiple libraries pooled together and I'm curious if you have done, or if you can do, some without pooling them. If the qPCRs work(ed), there's absolutely no reason for the sequencing to fail as any kind of molecular inhibition would/should affect the qPCR as well.

Lastly, can you isolate the ~50 reads from the tags that fail? Can you get them through an analysis pipeline for variant calling? What do they look like in IGV? Are the tags there and complete? The only two culprits I can think of at this point are things like binding to the flow cell and whether the sequencing primer binds correctly.

[GenoMax]

@vanessamata: I am not an experimental expert so take this with a grain of salt until someone confirms it but have you thought about heating your sample before loading? That should help break up any strange structures that may be forming. I remember this working for some difficult to sequence samples in past.

[Jessica_L]

That's a good point, GenoMax-- for the TruSeq Amplicon v1 that I've used, the library that's eluted off beads post-normalization is single stranded in dilute NaOH and that requires a short heating followed by snap cooling on ice prior to loading. It's not a bad idea to try. The Illumina protocol for the TruSeq Amplicon calls for 96 degrees C for 2 minutes, followed by 5 minutes in an ice water bath, if you wanted to try something similar.

[nucacidhunter]

Considering discussion so far the most likely cause would be errors in overhang primer design or synthesis such as a 3' single base mismatch which prevents priming sequencing. I would check sequences on the oligo tubes to see if it matches with the intended sequences. Also 5 out of 6 unsequenced tags start with T but it would less likely be the cause.

[vanessamata]

@GenoMax and Jessica-L

I think you might be right!

The 16S metagenomics illumina protocol also mentions this step! For some reason the technician responsible for preparing the libraries to load into MiSeq was not doing it!

Next time I will try to heat up the library and see if this still happens or not!

Thanks a lot!