Field Notes | ArcherDX Checks | It's Great for Fusions & Indels, Not Liquid Biopsy

Overview

We conducted three interviews with subject matter experts on ArcherDX and anchored multiplex PCR (AMP) from April 14th through April 27th: a former senior ArcherDX operations executive (>12 months removed), an experienced senior scientist at a Mid-Atlantic academic institution, and an Assistant Professor specializing in assay development at a Midwest academic institution.

We continue to wrestle with this one and felt like we needed to gain a greater appreciation for what ArcherDX does and doesn’t do as a part of our due diligence on Invitae ($NVTA), which remains near the top of our Watchlist. These calls were somewhat disappointing because rather than find a clear bull case, there is a gap between the utility of AMP that the company has promoted and what experts in the field described. Said differently, we're still struggling with the super exciting bull case that pushes NVTA over the proverbial line, and we expected to come away from these calls with a logical framework to model the revenue opportunity (boxes and throughput per box). We admit to being somewhat confused by the lack of enthusiasm and a seemingly narrow opportunity (fusion detection). NVTA is a MicroQuad 2 stock with accelerating estimate momentum (a great thing), and users do appreciate Archer's software, kits, and service level, so we're not giving up.

Takeaways

1. ArcherDX is "special" for gene fusion detection
2. For broader use - detection of insertions and deletions (indels) and blood-based ctDNA detection and analysis at the core of liquid biopsy applications - Archer's AMP technology is not required (i.e., there's no advantage to using Archer over others)
3. There are risks aside from competition - costs must be monitored closely and either the supply chain or IP could pose a problem and should be monitored as well

Interview 1 - Call Takeaways (Assistant Professor)

Our contact’s career has largely focused on the anchored multiplex method so he’s very familiar with ArcherDX, AmpliSeq an alternative, etc. The liquid biopsy angle gets a lot of attention because it’s a huge TAM; however, it’s a “suspicious story.”

Background of ArcherDX and AMP chemistry (target enrichment for NGS library prep): Where the “bread-n-butter” lies is in the detection of gene fusions, an actionable class of alterations that are most common in lung cancer. Outside of gene fusion assessment, AMP chemistry works fine, but it’s “nothing special” in terms of an application for liquid biopsy.

• Archer “isn’t doing anything fundamentally different than anyone else in liquid biopsy.”
• The error correction and molecular barcodes – that’s “not specific to Archer.” Those pieces are needed to reduce the signal-to-noise ratio, which is critical.
• The trick: looking for low allele frequency variants, you need a chemical and bioinformatic way to detect things at low frequency. You can’t under standard NGS. Guardant was the forerunner (some use of molecular barcode, some bioinformatics to drive limited detection low - where it needs to be) in liquid biopsy. 

Why is that important?

• There are cases where the tumor doesn’t shed much DNA into the bloodstream. For disease monitoring, you must be able to pick up on low amounts to make sure you’re not missing recurring disease. 5 years ago, I didn’t think it’d ever have legs – MRD that is. You wouldn’t want to give the patient peace of mind if it’s negative – does that truly rule out residual disease? What if there’s a growing tumor NOT pumping DNA into the blood. 
• That’s why I think MRD is still early. Guardant was the first to offer the assay and while I’m not sure how upfront they are about the chemistry or what they use in terms of bioinformatics, they get really low (mutation at 0.1% of the sample - they were the first to get there, then Foundation and others). 

• The trick of Archer, primer 1 to the gene, primer 2 is to the adapter that sticks on the fragment. Avoid missing things - not deleting a primer binding site. 
• Archer does employ a molecular barcode, as does everyone else.

• They sell great kits, lyophilized material. They’ve “mastered the ability to send kits out.”
• Labeled RUO (research use only) – but if you send to a clinical lab, they will validate the assay. In terms of direct application of AMP -> liquid biopsy, it’s not as special.
• Fusion could be in the mix, but it will be as RNA, which is the stage of translation where an indel occurs. Blood based nucleic acid is cfDNA or ctDNA because it’s more stable. RNA isn’t as stable – “the body wants to chew it up and get rid of it.”

• Fusions - primarily lung. NTRK fusions are a primary example. This is the first pan-cancer, specific target. Lung is the poster child. It’s a rapidly expanding field.

• The story has been gaining traction over the past 5-10 years. Lung cancer - guide therapy selection via EGFR mutation. We have seen multiple new drug targets identified – there are at least 8 genes where we’ve found alterations and now patient standard of care therapy targets those genes. Now we can look simultaneously.
• Before, we did one at a time. With NGS we can look at multiple.
• Many assays are pan-cancer. But you add to the number of genes you must look at simultaneously. There are still fewer than 50 that are directly clinically informative (can take the number up w/ trials). 
• Tumor mutational burden guides immunotherapy use. To get TMB - Foundation, Caris, Tempus, others - 100s of genes allow you to drive the score, but most won’t be informative.
• Is there a menu? How expensive is that? If you ask 100 people, you’ll get 100 different answers, but there’s a core set of ~20 important ones. With the bigger ones, the people that make assays are trying to be forward-thinking – i.e., maybe a marker will matter 5 years down the road. For example, passenger mutations are only that one patient, but it could be something to track for RMD.
• As sequencing gets cheaper and there are more things to do w/ the data that comes back from whole exome or whole genome sequencing, it could change. There will be several hundred that directly inform drug selection. Foundation has the most commonly used assay – it becomes expensive at $3,700 per (there are 350-ish genes on that assay).

• No, Nanopore for longer reads but in terms of this approach for MRD or liquid biopsy, ILMN is the player. Ion Torrent is a competitor but has issues you can’t get around. 

• When talking gene fusions it’s an example of the flaws. The set of opposing primers - they can’t detect novel gene fusions. That’s what Archer truly helps with.
• The way the market is set up, Foundation now has liquid and Guardant has tissue, but Foundation is still the leader in tissue. Tissue still needed for diagnosis - disease resistance is where liquid biopsy is helpful… if you become resistant, we want to know why. A lot of that is based on mutational analysis and it’s easier to do a blood draw. The tumor is likely shedding - specificity good, but not perfect. Sometimes there’s not enough tissue.
• It’s hard to bring liquid biopsy in-house because Guardant does a great job, Oncologists are happy with and like them.

• Of all the labs – the bigger academics likely do/use the gene fusion assay - ⅓ maybe. It’s far less for mutational/liquid biopsy.
• Is your use going up? I’ve/we’ve been heavy users since early on, but it’s been growing with drug approvals. They are doing something unique for gene fusion assessment - not the only way you can do it. Illumina has a kit that gets around some issues, but Archer is in the lead.

• We were down during COVID-19 (referrals were down from the community), but volume is quite high now (our contact couldn’t quantify it). Volume is definitely up. We never used to do GYN malignancies, but now it’s the standard of care for certain cancers (e.g., endothelial). So, volume exploded because of that, and then there’s a bit of a post-COVID surge.
• We lost a year of tracking patients, but cancer didn't go away. It’s anecdotal, but I think we’re seeing it now because volume is “way higher than it was” this time last year and certainly pre-COVID (ramp started late in 2020).

• It’s criminal that molecular is not standard of care, yet. More disease types appreciating that it’s vital information, and it’s spreading into the community (perception improving). Combine those and you see molecular becoming the standard of care.

Interview 2 - Call Takeaways (Senior Scientist)

He has a ton of experience w/ NGS in rare leukemia – every kind of sequencing, every technique. Last year, he took a technical role at an academic institution – helps with data out of panels like Archer or others. Biostatisticians run it for him. The goal is to identify real variants vs/ “garbage” – find anything actionable.

It was surprising to hear how manual curation still is - every data point coming off the sequencer. Some are screened by algo, but anything that makes it through is done by hand and requires a Ph.D. to recognize variations in terminology.   Archer’s AMP – one anchor allows for targeted resequencing. You can pull down what you’re interested in and save money. Targeted resequencing is cheaper if you want targeted panels. The whole genome, germline variants – depth of coverage and amount of data – the amount of data is overwhelming (10-fold difference at 300x coverage).

AmpliSeq (Ion Torrent) is the alternative to AMP (ArcherDX). There are others, but the main thing with targeted sequencing is the amount of material you’re starting with – “it’s always a problem.” To get what you want, PCR amplification introduces error - 40 cycles of PCR to get enough library to put on the instrument.

With AMP, in stage one of PCR, there’s a unique barcode under it. That captures it – one piece and you amplify it and can track it forward.

The whole point of everything after is you can look back and know what you started with. Unique molecular barcodes are NOT proprietary. But the barcoding that Archer uses allows for two bits of information that tell you what came from one capture, not just one PCR amplified X number of times, which is powerful for RNA. Even with Archer, 14-20 variants will come out of a typical patient’s sample. “~4 are usually real, and we must screen out up to 15-16.” You start with a smaller pie, so to speak. “The key point for DNA is that you know what you’re sampling at the beginning is true – you know it’s not a PCR duplicate.” This allows researchers to look at rarer variants (under 5% = rare).

There’s a lot of garbage, and ArcherDX is helpful with Indels (insertions and deletions), which have their own error profiles and are harder to figure out.

Will Indels be consistent? Are there variations?

• Structured variations can grow. Think about tumor-suppressing genes – put something in or make it shorter. Acute myeloid leukemia – get a cluster. There are some good single cell papers that go into different mutations – RAS for example. Indels stable enough, see a patient a year later, same Indel.

• I sit directly under the director – NGS is ordered by pathology as much as a clinician when we’re unsure of what something is. In some cases, it’s helpful to get a diagnosis as much as it is for treatment.
• How long does it take? If a report is negative, it can take less than a minute. If positive, it depends – from five minutes to hours (depends on whether it’s a novel mutation or something we’ve seen before and annotated).

• There are other ways to get you online – ArcherDx – they have the ability to make a targeted amplicon well, it’s the ability or willingness to do it. Other people build solutions, there are other ways to do it.
• For MRD, the thing to figure out is how to target and have a flexible product. He doesn’t know much about Natera’s selection process (nobody does), but the other thing for MRD is an ability to prove that there are still variants remaining. They may be using passenger mutations. Also, reimbursement for X number of mutations – what will insurance cover is a big question (i.e., what’s the X).
• Tracking back on the RNA piece is a benefit of Archer, but it’s more expensive. Universal panels – the amount of material and overall expense is too high to have it be routine.

• It’s a hard calculation. Hard to guess. UMIs or molecular barcodes, yes – Washington University, for example. Dual primer PCR amplicon method – more difficult than doing single probe method that Archer is doing.
• Ion Torrent AmpliSeq – 10 years, low input, get answer, but two years ago, ILMN licensed it. Key there, they can design PCR amplicon panels (an algo to tell where to sit - that’s the value: give a list of 40 genes, make 1000 amplicons to play well together, and produce good sequencing data).
• Everyone else is at a slight disadvantage due to that ability to anchor in one place – they own that.
• If they develop additional rules for the data – liquid tumors, acute myeloid leukemia – but you can determine residual disease well in flow cytometry. The moment that
• Anyone doing differentiated NGS in-house could have one – or if you need to know MRD in liquid tumors.

Data analysis and interpretation piece – the software – use dozens of times each week. There are some UX cleanup things. He knows what is being presented and it’s “very serviceable, understandable, and well-documented, plus there’s a helpful help document - it’s fully developed, very usable software.” Labs that don’t have experience can pick it up quickly (even a lab that’s never done sequencing before).

Archer is helpful; “a good partner if you’re trying to develop something.” This hasn’t changed since the Invitae acquisition (a concern/risk – “people were terrified that Invitae would absorb them”).

One thing Archer could do well = NGS panels designed for myeloid tumors – He thinks there’s “room to spread out.” Depends on volume. Also, if post-resection gets adoption, that could make ArcherDX incredibly popular. “The amount done for MRD would outpace the amount of NGS being done.”

Interview 3 - Call Takeaways (Operations Executive)

Labs buy the kits from Archer and use mostly Illumina boxes; the magic, which [Archer] management realized early on, is the software. With NGS sequencing data, the big complaint is that the data are overwhelming. The way the UI works and data are displayed are great. The raw display does clinicians little-to-no good (clinical reference labs literally don’t want to deal with it).

• Archer is still largely reliant on Illumina, thus the press for the new sequencer from PacBio.
• They claim to be agnostic, but it’s being run on an Illumina box. 

Questions he has: The biggest risks, in his opinion, surround the IP (how they license the anchor method), and the supply chain (disruption of supply of the enzymes they use as well as the lyophilized beads - source and unique materials; there have been court cases and lawsuits around ownership and access). Also, he mentioned GMP certification for the labs – it has been several years, but they still aren't. It was a “VC environment” - more so than a “rigorous processes and systems” environment. “There’s a time and place for that great story - but do you have right personnel to become real company?”

• The ArcherDX team commercialized AMP technology, which addresses the bottlenecks associated with using NGS in translational research. The gene fusion assays utilize targeted enrichment chemistry and automated bioinformatic analysis to detect both known and novel fusions from RNA (NTRK, ALK, etc.).
• He questions the moat around AMP, although it’s not clear why. He mentioned Myers leaving for GenapSys.

Cancer opportunity is huge overall. There’s a need for more targeted therapeutics and companion diagnostics (CDx) to play a big part. There are, however, a few things to consider. Being able to identify which patients will respond or highlight why some patients do not is hugely important. Because of this, it’s easy to see why ArcherDX is valuable to Invitae. However, there’s often a struggle in the relationship between pharma and diagnostic partners (not necessarily unique to Archer. A) Logistically, pharma partner doesn't want to commit to a relationship until fairly advanced (Phase 3) B) There can be a time crunch for Dx company. C) Processes - Dx vs. pharma can differ - stability testing, reagent formulation, etc. are not the same as what the pharma side must go through, but pharma is paying the bills.