Monday, August 24, 2015

Do Helix's Numbers Work?

A number of efforts in the consumer genomics space have been attempted in the past, with 23andMe appearing to make limited headway and Knome not much at all.  I haven't been able to get any investment interest in my own concept, though perhaps that's because it was tongue-in-cheek (or tongue held out while panting).  Last week brought a big splash, with a new company Helix launching with $100M and three major players as backers: Illumina, LabCorp and the Mayo Clinic

Helix is being billed as an app store for genomics.  The model, as covered in a number of outlets, is to try to charge very little upfront and then repeatedly sell analyses to customers.  Sequencing will be an "exome plus", covering most coding regions and some genetic control regions.  In interviews, Illumina CEO Jay Flatley suggested that the initial sequencing might cost the company the $500, but might be triggered by requesting only a $30 analysis.  These analyses would be developed by third parties, and reports have suggested that Helix would take a roughly 1./3 cut of each test's price.  That suggests that Helix will have to succeed in selling $1500 in tests per customer on average just to break even.

That's a lot of $30 tests.  I don't doubt there will be a wealth of tests -- each one is expected to be very narrow in focus, but that's still a lot of tests. Some of those will be repeats of earlier tests -- a customer's genome data won't change but the knowledge base might very well change.  Others might be prompted by new circumstances -- a customer has a new health issue driving new worries.  Sadly, it's very easy to imagine many will be the equivalent of genetic dowsing, giving dubious advice of skin care and diets.  Even if a customer orders over 50 tests during their lifetime that only covers the testing cost, not the cost of all that capital that was tied up.  Perhaps there will be some heavy users who order every test under the sun and run up off-the-charts billings, but while this model of works for gambling (and used to for texting), it's hard to see it quite fitting genetic tests.

I doubt Helix was started by dumb people, so how else might the company get money?  One obvious route is to sell access to the data (or digests of it) to pharmaceutical companies, as 23andMe has been angling to do. But, without detailed phenotypic data, it's hard to see much value in the genetics.  Will Helix be able to not only get consumers to trust them with their exomes but also detailed health histories?  That's a huge question.  Yet another route would be to try to push products to customers -- if genotype X suggests drug Y, would a Helix customer be given a starter coupon for that drug?

Finally, perhaps Helix will end up going strongly after healthcare providers.  Presumably they are going to have a sizable capacity for exome sequencing.  This speculation would look like this: a smaller medical facility (such as a community hospital) has its exome sequencing run at Helix at a better price than other providers, and in turn offers its patients a free initial account at Helix (perhaps with one "app" thrown in).  In a similar way, might Helix be able to entice 23andMe or to link up; the established players could simultaneously ditch their own laboratory infrastructure and offer more detailed analyses to their customers, while bringing a possibly more engaged client base.  Or perhaps Helix will aggressively pursue patient advocacy groups ala Patients Like Me, to try to get customers with intense interest in their genomes and health.

In any case, consumer genomics will be closer to mainstream. How many customers are ready for this?  That will be the real test: perhaps 23andMe has already taken the bulk of consumers interested in exploring this space.  Or perhaps future genetic news flurries, such as the recent one on an weight-related locus FTO, will give these services free advertising that drives sales.  Or a key celebrity endorsement, which for better or worse end up driving a lot of health-related spending these days.  But I suspect that many consumers will remain wary of this technology, particularly in this day-and-age where free services capture users but paid ones struggle.  Fears of genetic privacy won't be helped at all by some of the revenue mechanisms I suggest above, which will (fair or not) provoke thoughts of "big pharma" knowing you better than you know yourself.

On the regulatory front, I haven't seen much.  Will each app provider be responsible for negotiating the regulatory (and intellectual property) landscape of their tool?  That would be my guess, which should give any garage software developer some pause.  

So that $100M is going to be needed to pre-fund a lot of tests on the hope that they are later paid off, as well as designing an uber-secure storage system (especially after the publicity around the Ashley Madison hack -- which exposed some behaviorial phenotypes most people like to share only in a very targeted manner).  $100M sounds like a lot, but that could at most seed at most a hundred thousand or so tests -- and one should expect developing the infrastructure to cost several tens of millions of those dollars.  Biotech investing has been a net negative over time, and Helix is likely to spawn red ink for a long time.


Rick said...

Nice analysis. In particular the regulatory issues may prove to be a big tripping point as 23 And Me found out. It does seem like a financial model which relies on a large up-front cost being recouped by a bunch of smaller payments later is not a sustainable model.

Anonymous said...

Keith, sorry about anonimity, but this allows me to write freely.

I think you have fallen under an Illumina "illusion": the $500 exome. The reality is that Illumina's cost for an exome is actually closer to $50. And from what I understood (might be wrong) - they're still gonna charge $99 for the initial sequencing, which would be enough to cover costs and keep the exomes. If 23andme is valued at $1B with a 1M-people database of arrays, how much would a 1M-people exome company be worth? $5B? $10B? Plus, Illumina gains a very diverse control database and dilute their fixed cost on sequencing services.

Will it work? I don't know. Is it worth a $100M shot? Definitely.

Keith Robison said...

Anon: If your number is right for Illumina's internal exome cost, then it makes sense -- but that sounds very low since it must include initial DNA extraction. Flatley cited the $500 cost, which is why I used it (I don't have a good way to compute exome costs), but Illumina's actual cost could certainly be lower and perhaps Illumina will give them a really sweet deal on reagents.

An up-front fee would make more sense, but again that didn't seem to be how this has been pitched. On the other hand, a cheap entry means more people can be drawn in.

Will be interesting to see how this plays out.

Anonymous said...

I am very confident about my numbers, not so much their business model. At scale, saliva + DNA extraction will cost less than $5. Illumina's Nextera Exome list price is $100, and if you look at their quarterly statements, their profit margin is above 90%. That would make for a $10 prep. Then you have the sequencing, which costs (list price) about $500 on a HiSeq2500, $300 on a HiSeq4000, and $150 on a HiSeqX. Considering their margins are maintained for the HiSeq4000 (they probably have lower margins for X reagents), we have a $30 8GB exome. Add another $5 for bioinformatics (including 5-year storage) and you get a $50 end-to-end exome.

Illumina's fundamental business problem is their lack of price elasticity. If they reduced their profit margin overnight from 90% to 70% (thereby reducing list prices by 70%), they would take 2-3 years to see the increase in demand to get back to absolute revenue and profits at the current level. This means they are vulnerable not because competitors are close behind (they're not even close), but because they haven't found a moat to protect the castle.

It was once thought that bioinformatics could be a moat (reinforced by your next post - how ABI used to change their format often), but the dominant yet open standards stop this cold (thanks in great part to 1000 Genomes Project and the formerly-considered-open GATK). Google sort-of, kind-of tried to direct the standards in their favor with GA4GH standards, but it hasn't really caught on. Illumina is trying "data hoarding" as a strategy. Not confident that'll work either, but it could if we enter a Microsoft Windows era of slow innovation, but that's unlikely given Nanopore's quick evolution. Analysis in medical fields needs to be publicly validated and can be easily copied, so it's hard to build a moat.

In the end, Illumina is going to have a good run, but their strategy is full-stack Apple-style bundling. Makes for great profits as long as you have the best product, but won't allow them to act monopolistic. The question is who the Android and their Samsung will be to Illumina's iOS.

Brian Krueger said...

Came to say what anonymous said. And if you read between the lines, Keith, I would bet a billion dollars this is going to be a clinical exome product that only targets Mendelian genes so maybe a 10-15MB capture. That drives the total cost down quite a bit over a conventional 60MB+ exome (so like a $100 for those of us that have to pay markups and stuff :P). $50 sounds like a spot on cost to me in that scenario.

Anonymous said...

I guess investors were attracted by the Apple App Store Model more than anything in this case.. Its hard to predict if the numbers were right for the time frame that the illumina CEO is hoping for..