Wednesday, October 05, 2022

What Conversation Does Each Sequencing Instrument Vendor Wish You To Have?

In the TV drama Mad Men, a favorite maxim of lead character Donald Draper is "if you don't like what people are saying, change the conversation".  There's a strong nugget of truth to this: how we think about any topic is strongly affected by how the topic is framed - change the framing and often the conversation changes radically.  What might each sequencing vendor want us to be centering our conversations around, particularly with last week's unveiling of the Illumina NovaSeq X series?

Illumina: The World Consists of Human SNPs & Data Factories

Illumina's NovaSeq X series continues the upscaling of productivity with an upscaling of initial purchase price going with it -- and initial purchase price is highly correlated to ongoing service contract fees.  Illumina wants you to run lots of genomes, transcriptomes, spatial (as long as it generates Illumina libraries) and other -Seq assays.  That means huge sequencing factories (like the one I support) able to not only buy the box, but also buy and support the automation required to keep such a maw fed with libraries and the budget to pay for all those libraries.  It is somewhat surprising that Illumina hasn't offered a hardware product for high throughput library generation; apparently they are happy with the various liquid handling vendors competing in this space.  

The Illumina Complete Reads "no it's not a synthetic read, trust us" product is starting out aimed only at human and the emphasis -- particularly with the follow-on targeted approach -- seems to be just capturing some missed SNPs -- there wasn't much talk of SVs or haplotypes.  And if you target, you lose the ability to discover a lot of SVs and to resolve haplotypes.

Illumina would also like to talk about their built-in informatics capabilities, with DRAGEN showing up on all the latest boxes to perform routine FASTQ-to-VCF pipelines and more.  Having similar capabilities may soon be de rigueur in this space.  On the other hand, there's lots of applications not supported by DRAGEN, and some -- such as a lot around single cell work -- are compute hogs.  Don't be surprised to see such high profile, high compute code migrate onto DRAGEN  

MGI: We're Like Illumina, But You'll Save

MGI is eager to jump in and is the one competitor offering a full range of sequencers to match Illumina's, with pricing intended to give a discount on Illumina's offerings.  In theory the data is nearly interchangeable and MGI would like to present an end-to-end solution to customers.

What MGI would like to change the conversation away from is any perception that they are linked to the Chinese government or that supply chain could be at risk if a trade war broke out.  Those concerns have a degree of validity, but unfortunately it's easy for them to take on a xenophobic shade.

Thermo (Ion Torrent): Sequencing Innovation is Overrated; Our Customers Are Happy 

I haven't followed Ion Torrent closely for a long while.  The folks who used to eagerly reach out to me have scattered to four winds, plus every time I look at their website I don't see anything to encourage deep scrutiny.  There's several targeted sequencing niches that Ion Torrent supports well enough to keep their current customers, but they certainly aren't beating the bushes for new ones.

PacBio: High Accuracy SVs, Haplotypes and Methylation Are the Future

PacBio would like to convince you that you should be talking up accuracy of things beyond SNPs and short indels.  That getting methylation without extra steps is routinely useful from typical WGS samples - an open question in the WGS space though I'm told that there is some interesting data out there on pediatric neurological disorders.  That you should be looking for SVs routinely and be very interested in haplotypes.  They'd ask you to be skeptical of Illumina CLR's accuracy in calling these and point out it can't derive methylation and would say the input amounts (supposedly around 100ng for Illumina CLR, about 300ng for PacBio Low Input no-amp protocol) are close enough to not be an issue.  They'd like to chat plenty about the highest accuracy long reads in the business and point to the potential of DeepConsensus technology to push this even higher.

PacBio would also love to talk about the world beyond human genomes: high accuracy long reads have so many interesting applications for characterizing valuable agricultural germplasm, for synthetic biology & protein engineering and many others.  

Oh, and PacBio's going to get a new conversation going soon about their SBB product line, probably at ASHG late this month.  They're going to try to convince you that their improvement in read accuracy will translate into higher value, that the market is not sufficiently (or over-) served by Illumina read accuracy.

PacBio would prefer you not talk about workflow complexity and the dearth of automation solutions to support high throughput.

Oxford Nanopore: Our Data Quality is Really Good!  Believe Us!  And We're Light on Your Capital Budget

Nanopore is going to steer you to a similar conversation as PacBio around the glories of long reads, all the while convincing you that Kit14 data really is high quality.  And they'll talk up the much lower upfront purchase price, correspondingly lower service contracts and the lower maintenance demands of instruments with few moving parts.  Plus they don't take much space!

Oh, and also try to convince you that Short Fragment Mode is a worthwhile alternative to short reads.  Not that there's much publications to back that up, but trust them!

Nanopore would like to deflect any conversation around their clunky (and sometimes unusable on smartphones) websites and slow shipping that runs as smooth as a worn-out set of shocks.  Automated library prep is about as weakly developed as for PacBio, so they're not really worse -- but that's hardly a point to chat up!

Element Biosciences:  Illumina Has Badly Overshot Your Needs

Element has a commercial thesis that there is a large market for modestly priced boxes delivering around 1B (they just upped their spec) 2x150 reads with good quality.  I've seen chatter on Twitter to suggest a number of people are downright scared that their core will buy a NovaSeq Xplus, as they fear this will mean long waits to fill those mongo flow cells.  Charlie Johnson on Twitter enunciated a similar concern from the core lab operator's perspective: the instrument cost and service contract will demand a very high utilization or the core will lose money.  So there's certainly independent voices out there supporting Element's business plan. I would also expect there are many small biotechs that want the control and speed of in house sequencers, 

Element would also like you to chat about the value of immediate gratification - their box is shipping whereas the NovaSeq X is a minimum of six months away and the NextSeq 1000/2000 using XLEAP-SBS over a year away.  

Element also has the LoopSeq synthetic read technology, which they plan to leverage on their platform but have graciously still offered kits for Illumina.  

Element CEO Molly He on Twitter explicitly dared the sequencing kibitzome (the space of all who natter on publicly about the topic) to predict their next platform advances.  Presumably an overall increase in density would be a goal to enable more reads with a similar cost profile on the same instrument.  Growing numbers of analysis functions in their cloud -- perhaps they'll partner with some of the bioinformatics workflow curation companies that have emerged.  Element thinks LoopSeq fills the role for >150 basepair reads, but I'm a big fan of just a bit longer for a number of applications that may be very relevant for some biotechs that might be interested in AVITI - and not so interested in the labor intensive workflow for LoopSeq

Ultima: Wouldn't You Like to Save Big Bucks? And Don't Discount Our Data

Ultima is proposing expensive instruments that would drop the cost per gigabase into the $1 range.  It's been suggested that NovaSeq X might be discounted by Illumina to key customers to around the $1.5/gigabase range (versus $2 list), but for really data hungry groups that's still a sizable price gap. One question for Ultima (and to a degree, for any non-ILMN company) is how many Illumina shops can they convince to run both Ultima and Illumina or better yet switch over.  There is the Broad which would like to have one instrument of every flavor, but how many more such sites want to deal with the logistics, training and other complexity of a mixed shop?

There's also the ongoing question of to what degree is Ultima data functionally interchangable with Illumina data, given the homopolymer issue and single end nature.  Ultima would like to convince you that their talented informatics team will deliver both basecalling and downstream tools to smooth the difference; others are skeptical.

Ultima launching at around the same time as NovaSeq X should add additional excitement to the competition.

Singular Genomics: Let's Discuss Patience As A Cardinal Virtue

Singular is mired in supply chain issues that will keep them from delivering instruments.  The truly virtuous will wait for Singular, spurning the Element AVITI and MGI offerings and certainly NextSeq.  Nevermind that plots on the nets show Element beating Singular on upfront cost and running cost

Genapsys: Isn't the Phoenix An Amazing Story?

Genapsys (now Sequencing Health) would like you to talk about phoenixes rising from the ashes.  They've just gone through an ugly bankruptcy proceeding and the assets were purchased by a subset of the investors.  They'd like you to imagine the core technology moving past whatever held throughput to near toy range (though if they had delivered that soon after founder Hesaam Esfandyarpour gave a giddy AGBT talk in 2014 it would have done well)

Genapsys/Sequencing Health would rather not talk about zombies wandering around without direction, nor the flat trajectory of the last sequencing company to be resurrected after going bust, Helicos now SeqLL.


At some point I may try to survey sequencing startups again to look for signs some might emerge.  Roche/Genia/Stratos remains a closely guarded enigma, though both job ads and someone I talked to says the pot is still bubbling.  If Agilent (killed LaserGen) or QIAGEN (killed their own GeneReader) decided to jump in again -- or someone new to the space wanted to make an acquisition (before GE blew up I often wondered about GE Healthcare), Singular Genomics would be a candidate but realistically any of the startups could be.  Others have suggested that a major sequencing health care service might lock in a company by acquiring it -- I hope not (hate to see tech siloed) but it's certainly not out of the question.  Certainly at any moment the news could go out that there is a new player and they have a very definite conversational direction they'd like the world to take.

I Like Conversations Too!

As always, did I miss something?  Misconstrue something?  Join the conversation!

P.S. in touting the write-ups of NovaSeq X  / Illumina CLR by friends, I inadvertently neglected Shawn Baker's and since I wrote my piece Stephane Budel  and the DeciBio crowd also had a nice piece.  


Super Blogman said...

Hi Keith,

The link to Shawn Baker's blog should perhaps be ? You linked to a STAT blog by a different author, which also is relevant to this topic.

Anonymous said...

Ah, ah, funny article, and so true.

When it comes to the illumina vs MGI competition, and because you mentioned the Dragen, you could have wrote as well, in your illumina section, maybe they don't want you to compare the Dragen to MGI's Zbolt, because Zbolt is already faster than the Dragen for routine operations.

Anonymous said...

Really great post. Where does omniome fit in within the PacBio conversation?

Anonymous said...

As a beta tester, I can tell you Element's conversion kits do not work well for common library types and you can only run one flowcell at a time. I'm not sure how much flexibility will help Singular or MGI, but it is nice not to have to wait.

Anonymous said...

Nanopores main conversation point seems to be to get people to try for themselves and talk to other users.

Kristi said...

Anonymous: Your comment about Element is factually inaccurate (our sequencer can absolutely run two independent flow cells) and likely an NDA violation, as beta testers agreed to confidentiality.