Sophia Genetics: The Benefits/Drawbacks of Sequencer Agnosticism $soph

Akre Capital, Illumina/Ultima/Element, and Data-Driven Healthcare.

There’s no shortage of companies offering cancer screening tests. Grail’s Galleri screens for over fifty types of cancer from only a blood draw; Guardant’s Shield screens for colon cancer, Natera’s Signatera helps physicians track how patients are responding to cancer treatment etc. etc. For all these tests, a patient’s sample is sent to a centralized lab and sequenced/analyzed there. There are good reasons why it works this way apart from the FDA approval process: few doctors’ offices have sequencers on hand, sequencers are often expensive, they break frequently, and so on.¹ For the most part, it makes much more sense for Grail to develop its own test, pitch it to doctors, and then handle all aspects of the sequencing. Just as it’s not a good use of time for most companies to build their own payments infrastructure, it’s unlikely to be a good use of time for an oncology or primary care practice to operate a sequencing side hustle.

Sophia Genetics, a company I came across when looking at a recent Akre Capital 13F, predominantly serves a different type of customer: institutions that do have sequencing capabilities in-house and plan on keeping it that way. Rather than doing the sequencing, Sophia instead sells its software platform, SOPHiA DDM, to analyze those sequencing results using ML/AI (something it claimed to do before ChatGPT was released!) The company began, as many others have, offering oncology solutions (currently 70% of revenue), and has since branched out from there. There are two Sophia oncology applications that management has expressed particular enthusiasm about in the near-term: HRD testing, which helps inform treatment paths for solid tumors like ovarian cancer, and MSK-ACCESS, a liquid biopsy cancer test developed by Memorial Sloan Kettering that Sophia is commercializing.

Sophia existing primarily as a software provider that sits on top of sequencers has a variety of consequences, both good and bad:

·       It’s easy to branch out into additional forms of data. The company began with NGS data, and has since incorporated data from radiomics, digital pathology, proteomics, and others. It’s worth distinguishing Sophia’s multimodal efforts from those of sequencer companies like Illumina. Illumina management has emphasized its excitement around multiomics and customer appetite for sequencing more than just the genome.² That may represent a real opportunity, but Sophia’s is larger in virtue of existing as a data analysis company rather than a sequencing one. Pulling radiomics data into SOPHiA DDM is a growth avenue even if Illumina management is wrong and appetite for additional forms of sequencing doesn’t materialize.

·       Its revenue figures look weak when compared to better known testing companies like Natera, Guardant, or Grail. You can’t charge customers as much if you’re not running the tests yourself! This is partially why hospitals will choose to do sequencing in-house: they get to keep some of the economics.

·       The sales process is meaningfully different when compared to the above companies. Part of the headache of existing as a cancer detection company like Grail is the target market is theoretically every primary care doctor in the US. This presents a distribution challenge! Consequently, Grail has an interest in making it as simple as possible for physicians to order tests online. Sophia, however, has a sales process much more like a typical enterprise software company, which impacts both the selling timeline and the software setup period.

·       Not only are setup times longer, but selling software that analyzes NGS data means Sophia experiences additional delays when customers, or potential customers, decide to change sequencing providers. This was less of an issue until recently; Illumina remains the dominant player, but competition has ramped up significantly in the space from the likes of Ultima, BGI Genomics, Element, and others.³

Over the short term, elongated cycles and increased sequencer competition look like a headwind to the business. Over the long-term, however, there are upsides. The first is that, as would be expected, churn is lower than providers like Tempus. It’s easier to hold onto institutions than it is to hold onto individual physicians!⁴ The second is that a more competitive sequencing market is beneficial for Sophia for more than just the commoditization of its complement: If customers think there’s a good chance they’ll change providers in the next few years, they’re more likely to look for sequencer agnostic software rather than something like Illumina’s DRAGEN. Moreover, building software that works with different sequencers is no easy feat. Sequencers have different strengths, weaknesses, and forms of errors. The software used to analyze a sequencing run from an Illumina NovaSeq X system needs to be calibrated differently than software used to analyze a run from a PacBio Onso system (which is part of why sequencers are so sticky). Building out a platform that can work with any sequencer type is a real competitive advantage.

Having institutions as customers also comes with the more typical benefits of enterprise software: the opportunity to upsell. In Sophia’s case, that looks like offering additional forms of testing as well as expanding beyond the oncology departments of hospitals. Importantly, the SOPHiA DDM platform is free to access, with physicians only charged per analysis. The free-to-access model makes cross-department collaboration easy and an upsell more likely.⁵ Usage based pricing introduces more revenue uncertainty than a traditional software model (and was partially responsible for weaker Q2 results), but makes sense given the varying complexities of analyses performed and the healthcare reimbursement model. Reported net-dollar-retention figures are a bit all over the place: NDR came in at 142% in 2021 due to pent-up demand from lockdowns, but then was just 102% in 2022 given reduced Covid revenues and unfavorable FX movements (a downside of being an earlier-stage business but selling in multiple countries!) Notably, the 2023 figure came in strong at 130%, thanks to strong platform analysis growth for both new and existing applications.

Upselling also involves handling more of a customer’s sequencing workflow over time, which has the likely added benefit of reducing churn:

“We have experienced fluctuations in how our clinical customers access our SOPHiA DDM Platform across the three access models. Specifically, certain customers may transition from one access model to another over time. For example, we have observed a trend with certain customers being onboarded onto our platform through the dry lab access model, but, over time, as our relationships with them grow, these customers transition to the bundle access model as customers trust us to curate a set of instruments and consumable products to help increase the accuracy of the analysis they generate…..bundle access is typically a higher revenue-generating model compared to dry lab access based on the incremental value from the sale of consumables and instruments as well as higher platform usage on average for bundle access customers”⁶

At the risk of stating a tautology, the strength of Sophia’s data analysis software is dependent on the quality/quantity of the underlying data and the quality of its analysis. There’s reason to think Sophia has a real advantage here. Not only is the company able to use its customers’ pseudonymized patient data to train its various detection models, but the physician-in-the-loop model means doctors can flag genomic variants that they’ve found to be associated with a certain disease. These collective flags are then made available to other doctors on Sophia’s platform. Consequently, it’s not necessarily correct to evaluate the value of Sophia’s data only on the quantity and cleanliness of NGS/radiomics/digital pathology data points. Other relevant considerations are the number of doctors they can get to contribute to the platform and the quality of the doctors who contribute (hence it matters that Mayo is a customer!) The doctor in the loop piece has another added selling point: people tend to be less reticent to use applications that don’t seem poised to immediately replace them. This may especially matter for the institutions that Sophia counts as customers: a key part of Mayo Clinic’s pitch is that they hire the best physicians, not that they have the best AI models. Software that fits into rather than threatens this pitch has a better chance of being adopted by Mayo and similar institutions.

Sophia additionally operates a biopharma business segment that accounts for under 10% of revenue. Its biopharma solutions assist with drug discovery (leveraging Sophia’s vast amounts of data), clinical trial matching (leveraging Sophia’s large patient network), companion diagnostic development, and go-to-market strategy (leveraging Sophia’s data on geographical NGS testing trends for specific biomarkers). It’s fair to say execution on the biopharma side has been disappointing, particularly when you compare it to a company like Tempus. Having said that, there are a few different partnerships with AstraZeneca that are worth calling out. In 2022, the companies joined forces to expand HRD-testing for those with ovarian cancer in Spain. The rationale for this is straightforward: AstraZeneca offers a drug used to treat ovarian cancer, and that treatment is more easily covered by insurance if the patient has an HRD positive tumor. As a result, it makes economic sense for AstraZeneca to sponsor the HRD tests. The companies have since announced a second agreement to expand global access to liquid biopsy testing.⁷

It's a little challenging to know what to make of Sophia’s valuation. Like many 2021 IPOs, it’s down almost 80% since it first went public. Unlike 2021 IPOs, however, it’s not a business that was reliant upon Covid trends continuing or an abundance of venture-backed businesses to sell software to. It currently trades at a bit over 2x EV/’25 sales, a valuation that’s understandable if you think management slashing ‘24 revenue growth estimates from 25-30% down to 4-7% is indicative of a secular trend rather than a temporary blip. That said, there’s good reason to think growth will accelerate. Unless management’s lying, the primary drivers of its anemic revised revenue growth estimate were a weak EMEA market and unexpected challenges in its biopharma segment. The company’s executed well on its stated interest in focusing on the U.S. market, with U.S. revenue growing from 10% of the top line in 2021 to 15% in 2023 (after growing 70% YoY). As U.S. contribution continues to grow, EMEA weakness will become less relevant. I don’t know that I have a view on the biopharma piece, but I think the most important piece for Sophia in late fiscal ‘24/early ‘25 will be more signed customers actually contributing revenue:

“When a new customer signs up for SOPHiA DDM, it typically takes them about 9 months to begin generating revenue. This is because all institutions have to complete proficiency testing and receive accreditation before they can serve patients in routine. As highlighted during past earnings call, the 9 months sometimes becomes longer if the application they are adopting is more sophisticated like MSK-ACCESS. After completing implementation, it can take another 3 to 6 months for the accounts to ramp up to full volume and revenue. This means that new business won in Q2, Q1 and even some of the signings from Q4 2023 will not enter routine usage until late in Q4 or early 2025.’’⁸

I think this is the best argument for why Sophia’s stock is undervalued, particularly given MSK-ACCESS’ higher ASP.⁹ Fundamentally, Sophia’s still a software business with a strong land and expand model, a compelling data asset (particularly with its doctor in the loop piece) to feed its analysis products, and growth struggles that should subside over the next few quarters. The more the company can shift away from the EMEA market and towards the US, the more this business model will shine. Akre admittedly hasn’t historically made its money from growth-stage technology bets, but it’s a compelling signal that a fund with ~14.5 billion in assets and a concentrated style decided to spend time on a company with a market cap of ~235mm.

Disclaimer: The information in this post is not intended to be and does not constitute investment or financial advice. You should not make any decision based on the information presented without conducting independent due diligence

Disclosure: As of 11/01/24 I am long SOPH. Do your own research!

1. One of the (many!) interesting points from the FTC’s complaint against Illumina is how critical good customer support is given how often sequencers break: “Illumina’s instruments are ‘not like a washing machine…..[T]hey frequently stop working and you need to call an Illumina technician to come out and help find out what’s wrong with it and get it up and running again.”

2. This could include transcriptome sequencing, proteome sequencing, single cell analysis, etc. Growth in sequencing demand matters a lot for the Illumina bull case, as the fundamental bet is that delivering lower sequencing prices and expanded capabilities to customers will result in substantially increased demand.

3. As an aside, this means reading Sophia’s earnings call transcripts are a good way to get a sense for what the sequencer market is doing: “But definitively, while Illumina sequencers are the dominating one, we see more people adopting other types of sequencing technologies. In particular, I would highlight MGI [a subsidiary of BGI] and Element, we hear as well a little bit more about PacBio’s recent thing. And in terms of regions, I would say that maybe the U.S. is the one where, as far as of today, Illumina is still the one that is the preferred one. But in some other regions, we start seeing a little bit more competition” - Q1 ‘24 Earnings Call Transcript

4. Sophia’s churn for its Core Genomics Customers, defined as customers who generate revenue through bundle access, dry lab, or integrated access models, was a bit over 2% in 2023. Compare to Tempus’ metrics: “Through December 2023, the 12-month retention rate for physicians ordering more than 5 oncology NGS tests was 87% and for physicians ordering more than 25 oncology NGS tests, it was 92%. We define an active physician as those that have placed an oncology NGS test in the last 365 days.” (Pg 188, Tempus S1 Filing). That said, Sophia management did call out higher than expected churn as a reason for its lowered guidance this year.

5. This is analogous to Figma’s cross-side network effects, per Kevin Kwok: “Figma’s cross-side network effect offers an additional vector. Designers who use Figma share their designs with engineers and PMs, introducing them to Figma. As these non-designers learn to appreciate Figma, they then evangelize it to other teams of designers they work with on different projects. These cross-side network effects jump across teams and help Figma metastasize throughout entire organizations.”

6. Pg 93, Sophia Genetics 2023 Annual Report

7. Given that AstraZeneca is covering testing, the revenue isn’t considered biopharma.

8. Q2 2024 Earnings Call

9. “When it comes to the ASP, it’s [MSK-ACCESS] significantly much higher ASP because it’s a solution that is more sophisticated. And so indirectly, our average ASP will benefit as well from that as we were getting traction and utilization, which is something we expect to happen again from Q3, Q4 of this year.“ - Q1 2024 Earnings Call