[krobison]

I thought the average read length depended on whether you're using the 314 or the 316 chip.

This premise of this thread is a bit outdated, as the protocols have made several major changes since I started it.

I believe read length is independent of the chip (to first approximation); it is the newer protocols for preparing the template & better base calling software that are enabling longer reads.

[ningwang]

I'd have to really think through the dephasing, and probably model it.

Probability that any given molecule will dephase in a given cycle is the 1-repetitive yield (I'll call this Pdephase). If you are going to model this, one option is to assume that dephased reads are lost forever. Alternatively, one can assume they can be rephased under the right circumstances (indeed, Ion claims their flow order encourages recovering dephased molecules).

I'm not convinced that fewer molecules leads to less dephasing until you get to very small numbers. The current beads carry 800K molecules per bead; even with very high repetitive yield you are going to lose some in each round and presumably that is close to a fixed proportion. Clearly, if you get to molecule numbers approach or go below 1/Pdephase then you need to model it as a stochastic process. I'm not sure what Pdephase is, but presumably you could crudely estimate it from the distribution of read lengths.

If you get to really small numbers (but not 1), then a possible issue is that the relative contribution of a dephased molecule in the population becomes large -- so it is rare that a bead has any dephased molecules but a small number of dephased molecules kills your signal-to-noise ratio.

First of all, I don't know if Pdephase is dependent on how many molecules you have per bead. We know that PGM uses 2um diameter beads, which gives you a surface area of 12.5um2. If they're loading 800k molecules per bead, you've got roughly 15-16nm2 per molecule. If that dense enough for one polymerase to impede with another? If so, I suppose having a lower density might reduce dephasing. But if 15-16nm2 per molecule is ample room, having fewer molecules/bead won't change your dephased ratio.

Plus if you go with fewer molecules, at what point are you pushing the limits of the ion sensors at the bottom of the well? Also in the scenario you described (small # of molecules), you might mask dephasing problems but you'll increase your indel error rate.

[ningwang]

Does de-phasing impose any limits on their read lengths?

This is my guess as well. If you look at the supplementary data submitted by Rothberg in his Nature article (from July?), the raw data gets pretty ugly after 60-80bps and requires some heavy phase-correction to clean up.

[ningwang]

I've been told that with the Ion Fragment Kit, the maximum input fragment size should be 150 bp so that with adapters the fragment is no longer than 210 bp.

I thought the average read length depended on whether you're using the 314 or the 316 chip.

[arolfe]

We've used the 200bp kit. It seems to work as advertised and increases the maximum construct size (adapters + insert) from about 230bp to about 330bp. I don't have any data about whether the dropoff in sequencing efficiency is as steep at the top of the size range as it was for the 100bp protocol.

The only problem is that there's no OneTouch 200bp kit yet, so you have to do the ePCR by hand for the moment.

[[email protected]]

Has anyone tried the 200 bp kit yet? Using the 100 bp kit on variable size amplicons we see a strong bias against the longer (~200) bp amplicons so that amplicons around 150 bp are over-represented by proportion. This will cause us many problems in analysing environmentally-derived amplicons. Does anyone know how life tech plan to get around this for longer read lengths?

[krobison]

Sorry, by lost I meant they no longer contribute signal but do contribute noise.

The flow order improvement wouldn't show up in the code. The idea is that if you follow base X with another X after a few other bases, you can catch up some reads. For example, imagine the flow order

1 2 3
A T A

Incompletely extended molecules after cycle 1 will be finished off in cycle 3 and resynchronized in round 3, unless of course the next correct base is T in which case the majority of molecules on the bead will have marched ahead.

[lek2k]

I'm going to go off memory on this one so please correct me if I'm wrong.

1. dephased reads may be lost for ever but the contributions they make exist. This is evident in a ionogram as the zeromer calls look a little noisy and still exist deep into the read after rephasing.

2. According to the guys in Ion Community, the flow order can lead to dramatic improvements. However, this was only a simulation. I have found it very hard to find in the code where they have taken advantage of the 32 flow redundant cycle. Anyways, I suspect the long read data set is a combination of chemistry, flow order and other flow parameters being tweaked. There was also a change in version of ion-Analysis that hasn't been released. If you look at the summary report PDF the signal incorporation plot looks very, very odd.

3. Fewer molecules means weaker signal, which means good luck trying to differentiate between a zero-mer and one-mer when things start dephasing. Also I agree, if you have a fewer number then if a few dephase that would be a large proportion of the total making it difficult to rephase and base call. Then again, there could be a biochemical law that says there is a relationship between strands and polymerase and reagents?

4. For each chip sequencing run, the carry forward, incomplete extension and droop fitted probabilities are listed in one of the files for each CAFIE region (13x13 wells).

[krobison]

I'd have to really think through the dephasing, and probably model it.

Probability that any given molecule will dephase in a given cycle is the 1-repetitive yield (I'll call this Pdephase). If you are going to model this, one option is to assume that dephased reads are lost forever. Alternatively, one can assume they can be rephased under the right circumstances (indeed, Ion claims their flow order encourages recovering dephased molecules).

I'm not convinced that fewer molecules leads to less dephasing until you get to very small numbers. The current beads carry 800K molecules per bead; even with very high repetitive yield you are going to lose some in each round and presumably that is close to a fixed proportion. Clearly, if you get to molecule numbers approach or go below 1/Pdephase then you need to model it as a stochastic process. I'm not sure what Pdephase is, but presumably you could crudely estimate it from the distribution of read lengths.

If you get to really small numbers (but not 1), then a possible issue is that the relative contribution of a dephased molecule in the population becomes large -- so it is rare that a bead has any dephased molecules but a small number of dephased molecules kills your signal-to-noise ratio.

[BBoy]

But that does not seem to be the limiter at this time, since they are driving read length up without changes to the chemistry or chip. Or am I missing something? It seems that longer-term read length can be helped by more sensitive readout what would enable smaller beads with fewer strands on them. Assuming that the probability for dephasing is sqrt(N) where N is the number of strands on a bead, things should get better linearly with bead size, no?

[krobison]

Does de-phasing impose any limits on their read lengths?

Yes. De-phasing is the ultimate bugbear for any clonal sequencing approach.

Note that since my original query Ion has released a much longer dataset, though that protocol won't be public until sometime this fall.

[BBoy]

Apparently, the issue is the micelle size – material larger than 250bp doesn’t drive back to bead well. I suspect that this will no longer be the case with the new One Touch, since there is no oil in the system. At this point, that seems to be the only way to get to longer fragment lengths.

Does de-phasing impose any limits on their read lengths?

[Claudia Stewart]

I've been told that with the Ion Fragment Kit, the maximum input fragment size should be 150 bp so that with adapters the fragment is no longer than 210 bp.

Can anyone with access to Ion's deepest, darkest secret documents (known elsewhere as "package inserts") confirm this? The application note on Amplicon Sequencing mentions nothing on the subject.

Anyone know if the constraint is imposed in the library prep or the emPCR template prep? What imposes it? Extension times? Nucleotide concentrations? emPCR micelle sizes?

One might guess this is something Ion will need to address in order to hit 454-like read lengths

Apparently, the issue is the micelle size – material larger than 250bp doesn’t drive back to bead well. I suspect that this will no longer be the case with the new One Touch, since there is no oil in the system. At this point, that seems to be the only way to get to longer fragment lengths.

[zhengz]

this would be my guess as well. High copies per sphere until their algorithms get better?

As Ion prioritizes quick turnaround time, amplicon length has been sacrificed. The advertisement says read length is going to be double in 201x. It has been quite long time since 454 advertised 1000 bp, but still not there yet. Read length is the hard part for emPCR/bridge-PCR based technologies.

[ECO]

this would be my guess as well. High copies per sphere until their algorithms get better?

Also sensor/chemistry improvement perhaps? Algo can't detect protons that aren't there from a low level of template.