Amplicon sequencing

[epistatic]

Does anyone have experience with amplicon sequencing on the PGM to offer optimal amplicon length tips? I have reviewed their application note and the great information shared here and elsewhere.

My intent is to send off amplicons to see how the PGM works to validate SNPs discovered using the HiSeq on whole genome and whole exome samples from paired tumor/normals. I then want to screen many samples using the amplicon set.

The application note suggests 75 bp between primers which would give a final library size of 170-180 bp. This should take into account 100 bp read with - 4 bp for seq key and - ~20 bp for primer sequence to give 75 bp of informative sequence. Has this range worked well for other users?

Forward primer 30 bp + template specific primer 22-25 bp + Reverse primer 23 bp + template specific primer 22-25 bp = 97-103 bp (effectively 100 bp).

Is anyone close to publishing data using amplicon sequencing on the PGM?

[krobison]

I've posted my tale of woe when going above these size limits, but even so we called a bunch of variants from the data. It is definitely doable.

Depending on your primer design conditions, you may find that you can't design an amplicon with these restrictions, but I found that to be relatively rare. Also, you need to account for many of the reads not being so long -- the various public E.coli datasets can give you a real appreciation for what read lengths you will encounter. My own estimate is to try to design for 70-80 after the 4bp sequence key. A lot of this depends on how much you will try to pack the amplicons in, and also how much slop you need to leave for pipetting & amplification variation.

Example: suppose you expect to get 50K reads from a 314, being a bit pessimistic about how many are 80 long as well as chip yield (which varies). You really want a depth of 20 per amplicon, but shoot for 50 to allow for ~2X amplification/pipetting error. So you can then pack 1000 amplicons on one 314 run.

Of course, you might barcode your samples, but then need to feed that back into your length calculations.

[epistatic]

Do you see any reduction in sequencing yield with amplicons versus shotgun libraries? This is the norm in our 454 Jr data, I did not know if the same translated to the PGM.

[krobison]

I don't have data on that, plus our amplicon run was so screwed up by the size issue.

I've always wondered why amplicons have this issue on the 454 platform -- anyone have a clue?

[TomorrowIsToday]

You don't have to be limited to very small amplicons. You can amplify >400bp per amplicon. Then pool your different amplicons together and ion sheared using the Ion Xpress Fragment kit and make a library to sequence.

[RCJK]

I don't have data on that, plus our amplicon run was so screwed up by the size issue.

I've always wondered why amplicons have this issue on the 454 platform -- anyone have a clue?

Are you referring to reduced sequencing yield (lower # of reads) from amplicon runs? If so, I think a lot of this is attributed to the more stringent amplicon signal processing pipeline. By default it will discard an entire read if it is below a certain quality threshold, as opposed to trimming the end of the read until it meets the threshold as it would for a shotgun library. In some recent documents regarding amplicon sequencing on the 454, they describe how to change the filter settings to allow amplicons reads to be trimmed instead of discarded. This seems to make a big difference in some cases.

[krobison]

Are you referring to reduced sequencing yield (lower # of reads) from amplicon runs? If so, I think a lot of this is attributed to the more stringent amplicon signal processing pipeline. By default it will discard an entire read if it is below a certain quality threshold, as opposed to trimming the end of the read until it meets the threshold as it would for a shotgun library. In some recent documents regarding amplicon sequencing on the 454, they describe how to change the filter settings to allow amplicons reads to be trimmed instead of discarded. This seems to make a big difference in some cases.

Yes, I should have said "lower yield" and thanks for the explanation. For a lot of projects, trimming/marking would certainly be better than trashing.