amplicon sequencing on non-model organisms

[sfyoung]

I am soliciting advice on the real performance of benchtop high-throughput sequencers, particularly with respect to doing amplicon sequencing on non-model organisms at the scale of hundreds of individuals and 200 + amplicons. Our lab works on ecological applications almost exclusively with vertbrate taxa that lack reference genome sequences. We are shopping for a benchtop NGS intrument to do SNP discovery by RAD sequencing and genotyping by amplicon sequencing and have been in touch with Life Technologies regarding the Ion Proton and Illumina regarding the MiSeq. We cannot afford a HiSeq. We've heard the sales pitches, now we would like to hear from other labs that do similar work with non-model organisms about how the technology that is available now (spring 2013) performs in day-to-day use with respect to realized throughput and data quality. We would appreciate any information you can offer. For example,

Have you done RAD sequencing on the Ion Proton or MiSeq? How did it go?

Have you done amplicon sequencing with a custom panel of amplicons with the Ion Proton or MiSeq? How did it go?

Has the data quality from your instrument come close to what was advertised in the sales pitch?

How much specialized stuff did you need to get in addition to the sequencer to make the instrument work?

What is the least pleasant or most difficult part of the library preps?

Have you learned anything about your instrument that you wish you had been told before you bought it?

Would you buy the same instrument again (which one) or try an alternative (which one)?

Thanks

[Vinz]

We are successfully doing amplicon sequencing on the MiSeq with custom amplicon panels and custom sequencing strategy. MiSeq is a lot easier to handle than the PGM and gives good qualities. MiSeq has an issue with "low diversity". All short amplicon work is low diversity. However, they have work arounds in place (hardcoded phasing) and there will be a software update soon that overcomes the limitations.
We use qPCR for quantifying the library. That is basically what is needed in addition. We get 10 to 15 million reads on target routinely with 75 to 85% of base calls above Q30.
They are still having some issues with the stability of the instruments (in particular the valves). But they are quick on support and are getting better.
We would decide again for the MiSeq....

[MrGuy]

I actually find the NGS lab workflow for the instrumentation not important. Most of the time I spend is gathering the samples and designing experiments rather than digitizing them. A roche Jr could accomplish the same thing for my lab, it just costs more to run. The biggest benefit for me is that I can use the smallest chip (314), run it nearly empty, and still get all the data I need for the the amplicons designed. Anyone can do PCR, thus make a library. Amplicons vary from singleplex to very high (for manually designed) multiplex. There is no hacking of the instrument to resolve low diversity issues as amplicons are supported right out of the box. Our instrument hasn't had any service issues under the time we have owned it.

Do you really need so much throughput that you would require so many reads for RAD? Are you pooling or resolving down to the individual? What exactly is your design of experiment requiring for sensitivity? Necessary error rate tolerance may play a role.

Have you done RAD sequencing on the Ion Proton or MiSeq? How did it go?
*No

Have you done amplicon sequencing with a custom panel of amplicons with the Ion Proton or MiSeq? How did it go?
*Yes, went well and isn't very challenging assuming you are comfortable with optimizing high multplex PCR.

Has the data quality from your instrument come close to what was advertised in the sales pitch?
*Performs as expected. When it is garbage, we can trace it back to an error we made.

How much specialized stuff did you need to get in addition to the sequencer to make the instrument work?
*Nothing really. A qpcr isntrument (had it). The first runs we did without any special equipment aside from a qBit and egel. Everything worked just fine. We ended up getting some automation equipment later on.

What is the least pleasant or most difficult part of the library preps?
*input levels are still to high, generally speaking for NGS, in non-research oriented work. Libraries are actually easy today -- plenty of kits out there. No one company has a lock on a "simple" library prep as you can find an equivalent kit on the market (eg, Nextera/MuA Thermo)

Have you learned anything about your instrument that you wish you had been told before you bought it?
*The plug-in framework is pretty user friendly and allows us to customize analysis so "anyone" can be a bioinformatician for certain canned reports we've created.

Would you buy the same instrument again (which one) or try an alternative (which one)?
*Sure, why not. They are really all the same with minor nuances here and there. It mostly depends on what you are going to be using it primarily for. Amplicon? Ion. Whole Genome? Illumina. Both? Ion, possibly Illumina depending on the amount of amplicons run. Roche if I really only wanted amplicons and super long reads.

[brachysclereid]

MiSeq versus Ion

I am quite interested in the same questions as in sfyoung's original post.

Did you buy an instrument? Are you happy with the results?

[sfyoung]

We bought the MiSeq. We will do some SNP discovery and annotation and liked the ability the Illumina SBS platforms provide to work with paired ends based on flow cell coordinates rather than depending on alignments to a reference sequence. We work with taxa that do not have comprehensive reference sequences for alignments.

We have had the instrument for a couple months and are still evaluating the quality and quantity of data we get. Preliminarily we are happy with our choice but it is too early for us to brag about it.