[Yepler]

Just wanted to update (in case this helps someone else) and say that changing to the longer P5 sequence solved the problem. MiSeq returned an avalanche of good reads!

[jacksor1]

Problems with MiSeq Cluster Generation

Could you post the Illumina adpater sequences you had success with? I am having issues with cluster generation on our MiSeq usng PCR primers with adapters. I have found a number of references and they all seem to have different adapter sequences.
Thank you so much,
jacksor1

After several failed runs in a row (even loading at 50pM), I figured out my problem (with the help of very patient illumina tech support representatives ); so I thought I would update.

I had been using the old P5 and P7 sequences (single end versions, Nature 2008 paper; bentley et al; supp info) for the adaptors I PCR'd on. Apparently these changed for the paired end versions (same sequence with a few extra bases on the 3' end; see the nature 2008 paper).

So while my libraries worked for past GAIIx runs (because both primers were on the flow cells), at some point this changed and the MiSeq flow cells no longer contain these shorter P5/P7 sequences. Thus any clusters I was getting were due to very low levels of mispriming off the forward primer (reverse primer was fine since I am using standard multiplexing primers). Though those misprimed clusters would sequence fine.

Since all the quantification kits have the old P5/P7 sequences, all my RT-PCR titrations looked fine; it was just that the extra bases inhibited most cluster formation of my library.

I reordered the primers now, so I'll re-run everything again, but this explained all my previous problems.

Thanks everyone,
Sri

[Yepler]

Take a look at the supplemental info here:

David Bentley in Nature (2008) 456, 53-59 Accurate whole human genome sequencing using reversible terminator chemistry (supplementary material: http://www.nature.com/nature/journal...re07517-s1.pdf)

the adapter sequences I used were the ones described for the the paired-end flow cell. The ones used on the single-end flow cell were the ones I was previously using (and had failures with).

Hope that helps.

[nbogard]

Thus any clusters I was getting were due to very low levels of mispriming off the forward primer (reverse primer was fine since I am using standard multiplexing primers). Though those misprimed clusters would sequence fine.

Can you please elaborate on this? I am having similar problems with cluster generation on MiSeq but expect that my new libraries with PE adapters will work fine. My concern is whether or not the target for the sequencing primer is also abnormal. I'm worried that two low density runs I have completed (using single-read adapters) consist of useless and misleading data because, in my analysis, I only use the first base immediately downstream from the Read 1 sequencing primer (i.e. where the insert meets the adapter). In my case reads generated beyond that first base would be artifacts.

Do you think the SR adapters only cause a problem with flowcell binding? Would you expect sequencing priming to be normal?

[TonyBrooks]

We're getting similar problems, compounded by low diversity.
Our library is an pool of three different amplicons with three different R1 sequencing primers. We added some N's in our amplicon generation primers to immediately follow R1 primer annealing site in order to increase diversity during cluster registration. Apparently, it wasn't enough as our run was terrible (160k/mm2 density - Q-scores along the floor). We're going to repeat with 5% PhiX and the Josh Quick/Nick Loman fix (http://pathogenomics.bham.ac.uk/blog...illumina-miseq)

The issue I have is that several things worry me when looking at the thumbnails and general SAV analysis

1) Density is clearly higher than 160k/mm2 when eyeballing the thumbnails (poor cluster registration due to low diversity?)
2) There are a mix of low and high intensity clusters that can't be explained by laser cross-talk. These low diversity clusters are appearing on all channels (ATCG) not just the alternate base from the respective laser which is a bit strange, no? You can clearly see the same general cluster pattern for all bases - as if there's cross talk between all bases with some clusters lighting up fairly bright.
3) Only one amplicon seemed to generate any data post indexing

I'm wondering if two of the three read primers have too low Tm and hence generating lower signal. These are being filtered out leading to the lower cluster density estimate.
Tm's of each of the R1 primers are ~66 according to OligoCalc basic. Maybe I need to add some bases on to the other oligos? How do people check their Tm's? MiSeq runs at 65, if I'm not mistaken.

Comments and theories welcome!

[ECO]

We're getting similar problems, compounded by low diversity.
Our library is an pool of three different amplicons with three different R1 sequencing primers. We added some N's in our amplicon generation primers to immediately follow R1 primer annealing site in order to increase diversity during cluster registration. Apparently, it wasn't enough as our run was terrible (160k/mm2 density - Q-scores along the floor). We're going to repeat with 5% PhiX and the Josh Quick/Nick Loman fix (http://pathogenomics.bham.ac.uk/blog...illumina-miseq)

The issue I have is that several things worry me when looking at the thumbnails and general SAV analysis

1) Density is clearly higher than 160k/mm2 when eyeballing the thumbnails (poor cluster registration due to low diversity?)
2) There are a mix of low and high intensity clusters that can't be explained by laser cross-talk. These low diversity clusters are appearing on all channels (ATCG) not just the alternate base from the respective laser which is a bit strange, no? You can clearly see the same general cluster pattern for all bases - as if there's cross talk between all bases with some clusters lighting up fairly bright.
3) Only one amplicon seemed to generate any data post indexing

I'm wondering if two of the three read primers have too low Tm and hence generating lower signal. These are being filtered out leading to the lower cluster density estimate.
Tm's of each of the R1 primers are ~66 according to OligoCalc basic. Maybe I need to add some bases on to the other oligos? How do people check their Tm's? MiSeq runs at 65, if I'm not mistaken.

Comments and theories welcome!

The MiSeq Recipe for priming R1 is the following:

<ChemistryStep Description="FirstReadPreparation">
	<PumpToFlowcell ReagentName="LDR" AspirationRate="250" DispenseRate="2500" Volume="75" />
        <Temp Temperature="65" Duration="60000" />
        <PumpToFlowcell ReagentName="ReadOnePrimer" AspirationRate="2000" DispenseRate="2500" Volume="300" />
        <PumpToFlowcell ReagentName="ReadOnePrimer" AspirationRate="250" DispenseRate="2500" Volume="75" />
        <Temp Temperature="40" Duration="60000" />
        <PumpToFlowcell ReagentName="PR2" AspirationRate="2000" DispenseRate="2500" Volume="120" />
        <TempOff />
    </ChemistryStep>

[joshquick]

Hi Tony,

1) Agreed it looks about 2x that. lower reported density than you see in the Thumbnails could be caused by registration failures, you should see 0's in the P90 A/C/G/T if there was a registration failure in the imaging tab on SAV.
2) You expect to see some cross talk in A/C/T but G is clean, in your images G still has crosstalk which does suggest a mixed signal in the cluster - this would also result in a very low %PF which you see as it is a signal purity filter. Not sure what would cause this... Can you post the average matrix for the lane from Data>Intensities>Basecalls>Matrix>s_1_matrix.txt, it might give some clues.

The Tm's of the primers should be ok judging by your intensity plot (from @jamimmunology). If they are too low the intensity bombs between cycle 1 and 2 - your intensity actually looks quite good initially (C=1600) but something very strange happens around cycle 100.

Josh

[LSAD]

We encountered similar problem before (clusters on all channels) when we used sequencing primers annealing to P5/P7. Then we fixed the issue by using primers with similar Tm but annealing to other parts. What sequencing primers are you using?

We're getting similar problems, compounded by low diversity.
Our library is an pool of three different amplicons with three different R1 sequencing primers. We added some N's in our amplicon generation primers to immediately follow R1 primer annealing site in order to increase diversity during cluster registration. Apparently, it wasn't enough as our run was terrible (160k/mm2 density - Q-scores along the floor). We're going to repeat with 5% PhiX and the Josh Quick/Nick Loman fix (http://pathogenomics.bham.ac.uk/blog...illumina-miseq)

The issue I have is that several things worry me when looking at the thumbnails and general SAV analysis

1) Density is clearly higher than 160k/mm2 when eyeballing the thumbnails (poor cluster registration due to low diversity?)
2) There are a mix of low and high intensity clusters that can't be explained by laser cross-talk. These low diversity clusters are appearing on all channels (ATCG) not just the alternate base from the respective laser which is a bit strange, no? You can clearly see the same general cluster pattern for all bases - as if there's cross talk between all bases with some clusters lighting up fairly bright.
3) Only one amplicon seemed to generate any data post indexing

I'm wondering if two of the three read primers have too low Tm and hence generating lower signal. These are being filtered out leading to the lower cluster density estimate.
Tm's of each of the R1 primers are ~66 according to OligoCalc basic. Maybe I need to add some bases on to the other oligos? How do people check their Tm's? MiSeq runs at 65, if I'm not mistaken.

Comments and theories welcome!

[jme]

The sequencing primers are those used in the PCR for one end of the amplicons (25 or 26 nt), plus a different number of P5 nucleotides (either 6 or 9, depending on the primer) to bring the Tms in line with SP1.

My thought was that even if there is binding of sequencing primers to the lawn via their 5's, surely that shouldn't lead to fluorescence, as there's a 25/6 nt 3' overhang, so nothing for polymerase to act on? Although I can see how this could reduce the effective concentration of primer in the mix.

We actually can't move the primer any further into the amplicon, due to the variable nature of what we're sequencing, although I'm now toying with using LNA modifications to bring the Tms in line/above the 65°C mark without increasing (or maybe even reducing) the number of P5 nucleotides used. (Anyone have any recommendations for LNA sequencing primer design?)

[liepa]

We experience a problem with MiSeq in cluster generation (or cluster detection). We sequence PCR product of a certain gene and the P5 and P7 sequences for the adaptors and dual-indices are added during PCR (pair end).
We pool approx. 350 samples in subpools (with equal conc. of each sample), run on the gel, gel purify, quantify once again and form final pool (with equal conc. of subpools). The final pool quality and quantity looks perfect (for the last run we have purified even twice: 1. gel purification; 2. AMPure PCR purification). The subpools and final pool quantity has been estimated with Agilent 7500 DNA Chip. The first time we loaded on MiSeq 12pMol and the number of clusters formed was 188 with an output of 0.5million reads. We have rerun the same library loading 60pMol, which resulted in 1200 clusters and 5.2G sequences.
Illumina support refuses to help us, as it is self-designed primers.
We have reconsidered all the possible and impossible reasons and looked library preparation and quantification protocols numerous times. So I decided to inquire here, maybe somebody has experienced similar problems or knows what could be the possible explanation for this? Help or insight is really appreciated.

[microgirl123]

60 pMol is massive! There is obviously something very wrong if you are having to load this much sample to obtain good cluster density. I would recommend starting by quantifying your samples with qPCR to get a better handle on what is actually happening.

[liepa]

60 pMol is massive! There is obviously something very wrong if you are having to load this much sample to obtain good cluster density. I would recommend starting by quantifying your samples with qPCR to get a better handle on what is actually happening.

But do you think that we are underestimating the amounts we are pooling?

[microgirl123]

I'm not sure what I think, but I suspect that either 1. you are not loading as much DNA as you think you are, or 2. your adapters are somehow not right so that your sample is not hybridizing or not priming (you are denaturing before you load onto the MiSeq, right?).

The problem with using the Bioanalyzer to quantify your samples is that it is reading all the DNA present in your sample, whether or not the adapters are attached and whether or not the adapters are correct.

[memento]

I always use the following formula and it never let me down. It considers that the complete length of Your library (on average) is ~350bp (~250bp insert + 100bp adaptors). After pooling of samples (purified using XP beads or alike), I measure concentration using fluorometric assay (qubit or picogreen, high sensitivity).
I adjust the pooled library concentration to 1ng/ul. This gives me a library that is ~4.4nM (considering 350bp average length of pooled library). I take 0.2M of NaOH and mix 1:1 (v/v). This gives me concentration ~2.2nM and 0.1M of NaOH. Next I use hybridization buffer and dilute the library further, 1:150. This gives me a final load of my library ~15pM, NaOH is below 1uM.

NaOH is essential to make the library single stranded (unless You have some adaptors with forks that can be easily separated using 95C incubation/ice and immediate load). The final, low concentration (illumina suggest <1uM) of NaOH is also eseential since it can hinder binding of Your library to a flow cell. Using hybridization buffer to make final dilution is also a must. If You would let us know what is Your protocol right before loading it on machine (with timing for each step) I think somebody might point out some weak spots. I found qPCR methods very disappointing (maybe due to the fact, that I used PhiX dilutions as standards...).

[DRYTCYV]

Do you already tried the KAPA kit (qPCR) for library quantification? With this kit you quantify only fragments that have the adaptors for the flowcell.