Tuesday, January 12, 2016


Okay, third (and last) post for tonight.  Time for the victory lap -- and a reminder of the limits of educated guessing.  On Sunday I threw out a prediction of a hypothetical Illumina MiniSeq based on ordered arrays, NextSeq chemistry and optics, but with only one optical unit and a price point of about $50K -- and today Illumina announced precisely that.
Okay, not to get too self-congratulatory.  I don't know the insides of Illumina systems well, or I would have predicted the simplification from one pump to two.  It was only a conversation with Josh Quick that made me aware of 8 optical units in the NextSeq. $50K was a natural guess, as the MiSeq goes for $100K so any sequencer slotting in below it had to be a significant step below but ultimately optics and microfluidics cost something, However,  Illumina could have released an ultrafast system near MiSeq pricing that simply charged for speed.   I didn't put in a prediction for the data output, though I did think about it -- and would have been horribly under.  MiniSeq in high output mode has the same range of number of clusters as MiSeq v3 chemistry; I was thinking perhaps 1/2 or 1/4 as many.  And by the way, I will affirm that I had no inside information, no tips, no meetings in dark parking garages with shadowy informants -- this was simply an educated guess that came close to the mark.

MiniSeq is a bit faster than MiSeq, presumably primarily due to scanning only twice with two-color chemistry.  MiSeq v3 2x75 chemistry clocks at 21 hours but 2x75 on MiniSeq is only 13.  1x75 on MiniSeq cuts that time to 7 hours, similar to the time in Quick et al with tweaked MiSeq protocol -- but that imaged only half the flowcell.  This does suggest that Illumina could make further trades of output for speed.  2x150 is available in both the high-output and mid-output kits, but with 24 vs .17 hour runtimes.  The mid-output kit still generates over 2Gb in that manner, which could easily be far in excess of what some assays need.  

Its also interesting that Illumina is claiming all-in costs for sequencing of only $200-$300, which would seem to potentially make MiniSeq serious competition to MiSeq for many applications, cannibalizing the older instrument's market (though my perception of list prices for MiSeq may be off, since I pay via providers). One possibility of the current close spacing in performance between MiSeq and MiniSeq is a future makeover of the older platform.  Alternatively, MiSeq might remain only as the model for regulated spaces, and MiniSeq take over MiSeq markets.  For example, why is the longest chemistry on MiniSeq only 2x150?  Could 2x250 or 2x300 be supported?  Of course, I was just griping about issues with MiSeq 2x300 chemistry, and someone on my Twitter feed was averse to NextSeq 2-color chemistry for his particular scientific problem (though we've found it works well on high G+C DNA), so perhaps the close spacing is fine for now.  It would be unusual for Illumina to announce a new box outside of the current timeframe, but I wouldn't rule a MiSeq remodel mid-year.

At $50K, MiniSeq is half the price of MiSeq, but that's still a sizable capital item.  Physically, it's close to being a cube 50cm on a side (45W x 48D x 51.8H).  At 99 pounds and needing a standard 120V line, you aren't going to dream of backpacking one around in the field.  One person on Twitter claimed it was smaller than a coffee maker; I think they need to re-evaluate the ginormous quantity of coffee they are brewing and drinking.  Still, it is a benchtop machine that saves a bit of space over MiSeq.   

In terms of likely customers, the smaller size and cost could be very attractive to startups, particularly ones that are developing sequencing-based technologies or assays.  I wondered in the previous post how Illumina might compete with QIAGEN's carousel--like GeneReader platform, and perhaps MiniSeq is an effective solution.  However, if you wanted the freedom to launch a new assay every 2 hours, that's $600M list for the 12 NextSeqs (I haven't heard a price on GeneReader, so I can't compare).  As I mentioned in the previous post, many targeted assays are over-served by MiSeq, but the fact that MiniSeq has very similar performance doesn't really address this -- though perhaps with the reagent pricing Illumina is proposing nobody cares.

One more Illumina announcement to blog on, but it's getting late here & I'm almost as sleepy as Miss Amanda (who is curled next to me).  It would be apropos to blog on Firefly during the night, but I have caught Photinus in daylight on a few occasions, and I'll do so tomorrow.


Genetics said...

Nice. I read your little piece on a possible MiniSeq but didn't really think it would happen. And now here we are.....

Don't you love the MBA-marketing-experts and their marketing-(new)speak? If there are two output levels, nooooo, you cannot possibly call the one with the lower output simply the low-output kit, it has to be "mid-output" kit. "We don't sell anything low and inferior". As if customers buying an NG-sequencer are completely stupid and infantile. Reminds me of one of the first things in business I got told in the US: "In a meeting, don't talk about "a problem", that would make people VERY uneasy. Call it "an issue" and everyone will remain relaxed....".

What does Illumina say what is included in "all-in" costs?

200-300$ for what, 1x75 or 2x75 or 2x150, high- or mid-output? 200-300$ on the high-output (22-25 Mio reads) 2x150 would be quite something.

Not quite sure what the prices for competitor products in the US are, since I am from Europe, so I cannot really compare these 200-300$ against e.g. Thermo/Ion. But it looks like a product to fight off Thermo/Ion on the low-throughput NGS market (few samples per run, small panel-diagnostics, cancer-panel, microbiome). Considering that you really need an IonChef and a S5 to get comparable user convenience and considering what that will cost you and what Ion Chips and kits cost (at least here in Europe), the MiniSeq nearly eliminates the niche for Ion. What's left is a little time advantage, but running the Chef also goes over night usually.....
Ion consumables prices would need to go down substantially. Interesting market.....

And yes, the need for the current MiSeq seems to go towards zero with the new MiniSeq. If quality and read-numbers are what they claim for the MiniSeq, you only have the 2x300bp left pro MiSeq.



Genetics said...

Before posting my previous post, I didn't read the GenomeWeb piece. They say 200-300$ per sample, including library prep. Costs per sample are not very useful, of course. I had hoped for a "cost per run" figure.....


AKatawazi said...

So didn't they solve the issue with the NextSEQ with the v2 Chemistry, this is what Brian Bushnell posted for his results on V1 vs V2.

Hi all,

I have the comparative results now for the same library using NexSeq V1 and V2 chemistry, 2x150bp from a bacteria. V2 looks far better.





For V2 chemistry, the measured quality is much higher (particularly for read 1), the quality scores are more accurate, and the base frequencies don't diverge as much. Also, 79% of the reads mapped error-free, up from 50% for V1.

I also looked at HiSeq2500 data for the same library and it is similar in quality to the NextSeq V2.