MiSeq Reagent kit v3

[exo]

Hey all,

Anyone has tried using the MiSeq v3 chemistry yet? We just bought a kit and im planning a run next week, and i wanted to see what would you guys recommend concerning lib size selection/and loading concentration for an optimal cluster density.

Illumina announced that the optimal cluster density with the new kit will be around 1200-1500k/mm2. Using 4nM starting concetration, would that still be enough to get this much clusters? With my previous experience, I usually load 12pM and get a cluster den of 860k/mm2. You reckon if i load 20pM with the new chemistry that would be enough?

Best,
Elias

[LVAndrews]

Just started our first v3 run (600 cycles). I was expecting to have to load at a higher concentration to hit the higher cluster density, but after speaking with Illumina tech support I was led to believe that is not the case. The flowcells are identical to the v2 kits (except they come in a box that says v3). The reagents are the same except the one reagent is blue now instead of purple. Some of the reagents were modified to improve stability over the length of the run, but otherwise, I was told the real difference is in the software. The software will enable better identification of clusters which is what leads to the higher throughput. Not having witnessed the outcome, I admit I am skeptical at the moment, but I followed their advice and loaded as normal (18pM for these Kozarewa and Turner libraries). For v2 kits this usually puts me at ~900k/mm. Will see in a couple of hours.

I am skeptical since if the flowcell is the same, then in order to achieve higher density, one should load at higher concentration and then the software should better resolve the clusters resulting in higher %PF. I believe I have already observed an effect of the new software on the v2 kits since I am getting better %PF when I overcluster an have achieved PF output in excess of 10Gb with 500 cycle kits more than once now. Will post back tomorrow when I have run metrics.

[LVAndrews]

Loading at 18pM worked beautifully. The load concentration is probably off as I suspect the average fragment size is actually larger than the lab supplying the library has calculated, so please overlook that detail. Important thing is it works...

I was only reading bulletins about v3 chemistry and overlooked the faq describing the changes to the v3 process. Importantly, clustering and generation of run metrics takes longer (full metrics after about 5 hours rather than 3) which probably has something to do with achieving higher cluster densities without having to change loading practices. Anyway, this run clustered at 1650k/mm, 85% PF (32M), est yield 19.6Gb. Super impressed! After 140 cycles, the qscore heatmap displays a solid orange bar above q35 and I see about 95% of reads >q30.

I used the denature protocol that came out in March with 5ul 4nM pool + 5ul 0.2NaOH, 5min RT denature, and quenching with 990ul hyb buffer to make a 20pM denatured solution. Spiked in 5uL 20pM PhiX that I denatured last week to the 600uL load volume.

[Dun]

The flowcell might be different as well. Because there seems to be more tiles.
The V2 flowcell has 14 tiles each side,28 in total, while I found the v3 flowcell has 38 in total.
Any inside idea?

Just started our first v3 run (600 cycles). I was expecting to have to load at a higher concentration to hit the higher cluster density, but after speaking with Illumina tech support I was led to believe that is not the case. The flowcells are identical to the v2 kits (except they come in a box that says v3). The reagents are the same except the one reagent is blue now instead of purple. Some of the reagents were modified to improve stability over the length of the run, but otherwise, I was told the real difference is in the software. The software will enable better identification of clusters which is what leads to the higher throughput. Not having witnessed the outcome, I admit I am skeptical at the moment, but I followed their advice and loaded as normal (18pM for these Kozarewa and Turner libraries). For v2 kits this usually puts me at ~900k/mm. Will see in a couple of hours.

I am skeptical since if the flowcell is the same, then in order to achieve higher density, one should load at higher concentration and then the software should better resolve the clusters resulting in higher %PF. I believe I have already observed an effect of the new software on the v2 kits since I am getting better %PF when I overcluster an have achieved PF output in excess of 10Gb with 500 cycle kits more than once now. Will post back tomorrow when I have run metrics.

[LVAndrews]

I was told explicitly that the flowcell is the same. Since the software update for v3 chemistry improves the instruments ability to resolve clusters, it doesn't surprise me at all that you are counting additional tiles. This is probably the easiest way to go from 15M to 25M reads per run, and would explain the longer clustering stage and the longer cycle times since each would require imaging and processing additional flowcell spaces.

[microgirl123]

Our FAS told me to load at the same concentration as we do for V2 - the increased number of clusters is due to the MiSeq reading a greater area of the flow cell than with V2 chemistry.

[helico93561]

Just want to share our latest v3 600 cycle run experience.

the whole run took about 2.5 days.

we loaded 10pM DNA with insert size ranging around 1500bp, we have a cluster density of 700 with 69%>Q30.

estimated output of 10.7gb.

heat map showed weak signals after 250bp.

[MLog]

If the flowcells are the same for v. 3 and v. 2, this should mean that it is possible to get 25M reads using v.2 kit, isn't it? But this doesn't seem to be the case, unfortunately.

[exo]

Sorry just noticed this thread was active! So, I actually managed to sequence using the v3 chemistry last month. I just estimated that 20pM would cluster at around 1200-1400 - since previously, we clusted at 890 with 12pM.
And interestingly enough, cluster density was 1285k/mm2 with 89.8% PF. 60M reads with 53M passing filter and 85% with Q30 and more. Which was amazing!
So how i understand it, The flowcell itself isnt different but they increased the volume one of the reagents so that you can perform more cycles without running out. And the MAJOR difference is in the MCS, where in the new version you can now image 19 tiles instead of 14. I think what this means is that the MCS can now zoom in to smaller tiles, which makes it easier for the software to detect clusters at higher density. You could have easily clustered between 1200-1400 with the v2 chemistry, but the software would have had more problems with a bigger imaging area. Not sure if im making sense?

So loading the same concentration as you previously loaded with v2 doesnt make any sense for me. Estimate what works best for you to be able to cluster between 1200-1400, and the quality will still be great!
Best,
Elias

[LVAndrews]

Careful making corrections for cluster density. The flowcell is the same, but the number of imaged tiles has increased as mentioned above. Illumina support told me to load at the SAME concentration as normal for v2, and indeed it clustered higher as I was told it would. The clustering step takes longer, so I suspect there is a longer initial anneal step to increase cluster density. If you had to increase your calculated concentration to achieve the higher density, you have made a mistake somewhere.

[exo]

Well, previously i was clustering at around 800k/mm2. Thats why i had to increase my input concentration to hit the 1200-1400 range.
Can you say what was your cluster density with v2 and v3 using the same conc?
They clearly mention on their website:

http://www.illumina.com/systems/miseq/performance_specifications.ilmn

"†Install specifications based on Illumina PhiX control library at supported cluster densities (between 880-965 k/mm2 clusters passing filter for v2 chemistry and 1200-1400 k/mm2 raw clusters for v3 chemistry). Actual performance parameters may vary based on sample type, sample quality, and clusters passing filter."

Which clearly means that you can cluster at a much higher density with v3, and the cluster density should be only effected by input conc.

[LVAndrews]

Now that I have said that, I am questioning density once again. My previous experience has a serious caveat, that the lab providing the libraries does not correctly calculate insert size, so I have calculated that loading at "18pM" will provide the desired results. Loading at this artificial concentration, I was hitting 850-1100k with v2, and got to 1600 with v3 with no adjustment in concentration. However, since they are doing two-sided bead selection, their sizing could vary quite a bit run to run, which would explain the variability.

I have been running amplicons from our lab, and am very meticulous and there is no error in fragment size (16S v4 construct is 381bp). Loading at 5pM with amplicons seemed to produce consistent clustering between 950-1200 with v2. But yesterday I started a v3 run with amplicons, and it clustered at only 660k, which I might normally consider a low-end cluster density for v2, but should have been much higher for v3 in my opinion. I will consult with my FAS and post an update.

Sorry for the premature accusation of mistake making, as it seems I am the one doing such things at the moment.

[LVAndrews]

My FAS is out so I spoke with someone at sequencing support. They told me the recommendation for v3 is to load at the same concentration your first time, or maybe a little higher and then bump up as necessary to hit the higher density, so yes, you do need to increase your concentration to cluster higher. My previous impression from a conversation with them a month ago was that you should load at the same concentration as for v2.

Explicitly, I was told that to achieve optimal density with v2 (using phix), one should cluster at 12.5pM, and with v3 at 20pM.

I attribute my experience with the genome run (which appeared to cluster higher for v3) to possible variability in the size determination by the lab who provided it. My experience with the current amplicon run supports that clustering is the same on a v2 as on a v3 cartridge.

I will be rerunning the current library at a higher concentration in the near future.

[ECO]

Explicitly, I was told that to achieve optimal density with v2 (using phix), one should cluster at 12.5pM, and with v3 at 20pM.

We are getting about half cluster density on V3 with the exact same sample (split denatured pool) as we are on V2.

[LVAndrews]

Amplicons or genomic/other prep?