[DNATECH]

Some HiSeq3000 data are analyzed in this thread:

http://seqanswers.com/forums/showthread.php?t=58353

[frozenlyse]

Slight correction:
A HiSeq2500 1T should be able to do 500 GB per flowcell in 6 days. (83Gb/day) A HiSeq4000 can do about 700Gb per flowcell in 3.5 days (200Gb/day).

Still, a major increase!

--
Phillip

Which means that people will be doing more runs per year which means a lot more reagent sales for Illumina!

[austinso]

In the sense that a 3000 will produce about the same amount of sequence per unit time as a 2500-1T. Also in the sense that Illumina priced the 3000 at the current 2500 price.

But in the sense that the 2500 is a 2 flow cell instrument it would be replaced by the 4000 as a 3000 has only one flow cell.

And in the sense that the people who bought the 2500 in Q4 will be potentially be offered a 3000 and not a 4000. I'm entertaining the Illumina POV here...

But this doesn't imply that rapid chemistry will never be available on the 3000/4000, does it?

Correct, it doesn't, so yes I'm reading between the lines here...

I'd say given the push for the NextSeqDx, and dropping the HiSeq2500 for 510(k) clearance, it makes less sense to introduce rapid mode capacity (or even maintain for the 2500) for the 3000/4000 when there is a NextSeq500 that is being pushed as a product.

Besides, I don't think that ordered clustering is something that can be pushed in the same kind of way any time soon...

IMHO, of course...

[pmiguel]

2. HiSeq3000 effectively replaces the HiSeq2500...not the HiSeq4000.

In the sense that a 3000 will produce about the same amount of sequence per unit time as a 2500-1T. Also in the sense that Illumina priced the 3000 at the current 2500 price.
But in the sense that the 2500 is a 2 flow cell instrument it would be replaced by the 4000 as a 3000 has only one flow cell.

3. rapid mode will not be supported by either the HiSeq3000/4000.

I only see this from Jay Flatley to support your statement:

There will be a number of customers who remain on the 2500, however, due to the use of validated workflows or the importance of access to rapid run mode and longer reads.

But this doesn't imply that rapid chemistry will never be available on the 3000/4000, does it?

--
Phillip

[austinso]

FWIW...I think the Q4 results transcripts say a lot more about Illumina's product roadmap:

http://www.seekingalpha.com/article/2856246

1. HiSeq2500 is being removed from FDA clearance, and will be replaced with the NextSeq with NIPT.

2. HiSeq3000 effectively replaces the HiSeq2500...not the HiSeq4000.

3. rapid mode will not be supported by either the HiSeq3000/4000.

Bright shiny objects usually have a purpose...

[pmiguel]

- HiSeq 3000 is upgradeable to HiSeq 4000 (the HiSeq v4 upgrade was only available on instruments that shipped after 2013).

HiSeq 3000 is to HiSeq 4000 as HiSeq 1500 was to HiSeq 2500. 1 flow cell vs 2 flow cells.

The "1T" upgrade available to newer HiSeqs, that allows use of v4 chemistry, was independent of whether they had 1 or 2 flow cells.

--
Phillip

[Genohub]

Looks like most of the points on the latest Illumina tech have been covered in this thread. A few others points that I haven't seen covered yet:

- HiSeq 3000 will cost around the same as HiSeq 2500 v4.
- HiSeq 3000 is upgradeable to HiSeq 4000 (the HiSeq v4 upgrade was only available on instruments that shipped after 2013).
- You can upgrade from X Five to X Ten...just buy more instruments, but at the lower $1M/instrument cost.
- Illumina is going to start bundling library prep and cluster reagents on the X series.
- X series is still only for whole human WGS...likely to focus battle w/ the BGI / CG WGS instrument release later this year or a good excuse to release more patterned flow cell models....

We've written some more details about the new release here: http://blog.genohub.com/illuminas-la...-and-hiseq-x5/.

As a platform that regularly handles sequencing service listings and orders, Genohub collects a lot of information on price and changes in price once a new instrument is released. In general, we've noticed what others have posted in this thread, price per Gb is rarely more competitive immediately after an instrument is released. There are still many providers offering excellent turnaround times and prices on HiSeq services that will continue to be competitive even after we have our first HiSeq 3000/4000 listed on Genohub.

[pmiguel]

Increased throughput (reduced run time) appears to be the main thing. Would be nice if it translates to reduced costs but for that I think they need some serious competiton (BGI, not PII...).

Hiseq2500: 450-500 Gb, 6 days 4000: > 400 Gb/day

Slight correction:
A HiSeq2500 1T should be able to do 500 GB per flowcell in 6 days. (83Gb/day) A HiSeq4000 can do about 700Gb per flowcell in 3.5 days (200Gb/day).

Still, a major increase!

--
Phillip

[Ingeneious]

I know nobody here cares about microarrays (nor should we?), but I think Illumina deserves some serious mad-scientist points for the NextSeq 550. If I'm understanding correctly, they've made a microarray adapter that fits in the flow-cell slot, and are reading it with the same scanner. More innovation than I'd expect from a company with a secure monopoly.

Maybe this is an attempt to wean the last few stragglers off arrays and get them into sequencing?

They had the HiScanSQ that did arrays plus sequencing and that was quickly discontinued. Asked an Illumina employee about it and he said it seemed like they were trying to make peanut butter and jelly again - he couldn't figure out the logic in it either.

It's cheaper than the NextSeq 500 and iScan combined, so might be useful for research labs looking for a sequencer and array as well or those doing scanning as followups. Seems very niche though.

[Chipper]

Increased throughput (reduced run time) appears to be the main thing. Would be nice if it translates to reduced costs but for that I think they need some serious competiton (BGI, not PII...).

Hiseq2500: 450-500 Gb, 6 days 4000: > 400 Gb/day

[pmiguel]

The difference is even less if you compare HiSeq 2500 V4 to HiSeq 4000. The HiSeq 4000's specs claim 4.3-5B reads vs up to 4B reads on a HiSeq 2500. After factoring in instrument and overhead costs I would imagine that pricing for the 4000 runs may be higher for the first 6-12 months.

Yes, I think maybe the main benefit from the 3000/4000 over a 2500-1T running v4 chemistry would be less worries about overclustering a flow cell. And that is big--on a psychological level. But I think most cores have systems in place that either make this rare or just leave it up to the customers to send correctly titred libraries.

--
Phillip

[kcchan]

The difference is even less if you compare HiSeq 2500 V4 to HiSeq 4000. The HiSeq 4000's specs claim 4.3-5B reads vs up to 4B reads on a HiSeq 2500. After factoring in instrument and overhead costs I would imagine that pricing for the 4000 runs may be higher for the first 6-12 months.

[pmiguel]

No, I'm interested in total number of reads to calculate coverage.

So, what this (thread) is saying, is that for the same amount of money on a HiSeq2500 RR, if I did the run on a HiSeq3000 - I can potentially get >10x the amount of reads?

HiSeq2500 360M reads

HiSeq3000 4.2B reads

Have I got this right? If so, this is amazing!

No, at best you may be looking at 1.5X the amount of sequence per $. (That 4.2B reads figure is likely for an entire flowcell.)

My guess is it will be slightly less than this as Illumina always combines an improvement in instrumentation/chemistry with an increase in reagent costs.

Plus there is the cost of the new instrument itself that may contribute to the cost of the service.

--
Phillip

[andibody]

High Output.
True, it's more like 270-290M with Q>30.
Still, it's 'only' a 15% increase. (again, not countering in speed and evenness of cluster density).

[kmcarr]

There should be an exact max of clusters for HiSeq 3000/4000 defined by the number of microwells (325M as stated above?).
We're getting regularly 300M reads on the upgraded 2500. If 325M is correct, the increase is not very pronounced if the shortened run durations are not countered in.

Are you sure you are comparing the same numbers (read, clusters, lanes, flow cell)? And for the HiSeq 2500 numbers are you talking about Rapid Run or V4 High Output?

The HiSeq 3000/4000 specs are 325M clusters per lane, which would be 650M reads per lane for paired end runs (using Illumina's method of counting).

The HiSeq 1500/2500 specs are 250M clusters per lane for V4 High Output (500M reads per lane). Our experience is pushing even slightly past 280M clusters per lane results in dramatic decreases in %PF and read quality.