MiSeq paired-end sequencing of an amplicon with 45bp inverted repeat

[lcarey]

Hi,
I have a plasmid library that looks like this:

uniq-seq1___45bp__sequence-of-interest___45bp__uniq-seq2

The 45bp sequences are identical and oriented as inverted repeats. I'd like to sequence the middle region and maintain orientation (so using primers in the unique sequences that flank the repeats (green/purple)). Will this be possible on a MiSeq, or will I run into problems in cluster generation or sequencing because of the inverted repeats? I can PCR this amplicon without much trouble, so I'm not too worried about generating the library for sequencing.
The repeats are 67% A/T if it matters.

thanks for any help you can provide
-Lucas

[share]

Hi Luke

It is possible to do with MiSeq. Only thing you need to do is to add 4 degenerative bases after Illumina sequencing priming site. These. 4 bases help in cluster registration.

[lcarey]

I just got results from 1% of a paired-end 150x2 run, and at least there weren't major problems due to the repeats. I'm still looking at the data. For the MiSeq is 4 bases enough to generate complexity so that the entire run can be a single PCR product? I was under the impression that that more complexity was required for cluster generation.
-Lucas

[JamieHeather]

There's a good article about this on Nick Loman's blog.

Bottom line, yes ideally you need more base diversity longer than 4 cycles in, but you can also get this from spiking in PhiX or some other diverse sample.

[share]

The first four random bases help in two issues
1. Template generation (cluster registration) after cycle 4.
2. Cross talk matrix calculations which happens after cycle 2.
When it comes to focus MiSeq uses green laser which excites g and t so if your first four bases are randomly distributed then focus will not be a problem.
Phasing per phasing calculations happens after 12th cycle these values are depended on n-1 and n+1 DNA strands in clusters rather than your base distribution. Spiking 10 to 15% Phil helps in error estimations and obviously for little base diversity but it definitely won't help if there are no 4 degenerative bases.
If cluster numbers are too tigh Rta may struck in loop but still it generates proper data. We used this technique in our lab and the resultant are satisfactory.

[GenoMax]

You will need to keep an eye on Q-scores. Registration of the clusters can be easily solved by adding diversity but the quality values can take a big nose dive once MiSeq encounters consistent low-diversity sequences.

[GW_OK]

Random bases at the beginning is a good start, but while it helps the initial cluster calling the rest of the reads are basically identical and that massive signal kills the quality, as we've found out here. The most elegant solution I've come across was actually suggested by Illumina at a tech day I was at recently. Basically you make your primers but generate several sets with increasing amounts of randomness at the beginning. Something like this:

N-Tag-Primer-Target
NN-Tag-Primer-Target
NNN-Tag-Primer-Target
NNNN-Tag-Primer-Target
NNNNN-Tag-Primer-Target
...
N(n)-Tag-Primer-Target.
Where N is a random base.

This not only adds randomness at the beginning of the read but it moves the rest of the target amplicon out of phase for each respective sample, significantly removing the burden of the massive signal spike. The only downside is that it requires more than one primer set. You could probably get away with a few samples in each "phase" if you included the tag and made sure there was a relatively equal amount of each different phase. The tag itself could also add complexity.