The Future of Sequencing

This article was originally published in Inside Precision Medicine. But I thought my original draft might be of interest to readers here.

The first thing I imagine when asked to write about the future of sequencing… One of the various McFly children straight out of Back to the Future 2 coasts into the lab on their hoverboard. They skids over to the bench and press a single bright button labeled “Sequlite 10000”. A blur of motion and CG effects engulf the lab. Ten seconds later a AI avatar appears from some previously unseen corner¹ and ( proclaims “Paper Submitted and Accepted” (in the future all papers are written, submitted and reviewed by ChatGPT 10”).

When I woke up this morning, with the clarity of having slept on the idea, I realized that what I actually want is far more prosaic. I could talk about lab automation. Or sample prep innovation. Or long reads. Direct detection of RNA, methylation all the new exciting approaches that might exist… but then I started to think about what I actually want. And how most of those things are technologically feasible today…

Boring Sequencers like Boring qPCR Machines

What I actually want is for sequencers to be as boring and simple as a qPCR machine.

Nobody get excited about buying a qPCR machine. Nobody feels a tingle of excitement when they put that new qPCR machine (which is the same as all the other qPCR machines) on a grant. No… qPCR machines are boring. It’s stable technology that hasn’t really changed much in the past 10 years. You can’t just buy a new fancy qPCR machine and assume you’ll get a few papers out of it because everyone else has one already and they just kind of work.

And that what I want from the future of sequencing.

I want to be able to buy a sequencer from a vendor (or grab one from a used lab supply company) stick it in my lab and use it. I want to be able to buy the reagents and consumables from one of tens (if not hundreds) of suppliers and throw them on my old sequencer which may (or may not) have a service contract.

You can do this with a qPCR machine, in fact your lowest cost of entry is shockingly low. Buy an old Cepheid SmartCycler for $200, even a BioRad or ABI OneStep will only set you back a few thousand dollars and a new instrument can be had for $20K. These will all be fine for basic research applications. Similarly if you’re setting up a new lab, you can get competitive quoting from a number of different vendors all of whom will be providing near equivalent machines. Sequencers? Not so much.

Illumina dominates the market, and the competitors are all a bit different. The workflow is different, the throughput is different, the data quality is different. Competitive quoting is therefore somewhat of an artificial exercise… none of these platforms are drop in replacements for the others.

But they could be! And that would be pretty great! Imagine not being locked into a single vendor because you’ve built your whole workflow around them? Imagine being able to switch to another vendor because you don’t like the support you’re receiving, or the service contract they want to lock you into.

Imagine (shock horror) not being forced to buy your sequencing reagents and consumables for the same vendor as the instrument! Imagine (like qPCR) being able to choose from literally hundreds of vendors willing to supply you.

The fact is that there’s not technical reason why this world couldn’t exist today. The fundamental patents around next-gen sequencing have all expired. Building generic sequencing instrumentation is a realistic possibility.

However, building this new eco-system around next-gen sequencing requires a significant shift in the market. We can only dream!

More, Different, Automation

Automation around sequencing currently seems to take two forms. Space-age digital fluidic solutions which also cost about as much to a trip to the moon, or pipette robots that seem to do their best to copy exactly what your lab tech does. They costs thousands of dollars, are only really useful when processing large numbers of samples. They usually come paired with their own automation specialist tasked with keeping them happy, so they don’t smash themselves (and your samples) to pieces.

I’d like to see a very different kind of automation. One that is barely recognizable under that name. Taking another leaf out of the qPCR playbook I’d like to see sample-to-answer automation. The Cepheid GeneXpert is a good example of this in the qPCR space. Here a user can pipette in a raw sample which is processed through a fixed workflow (extraction, reverse transcription and finally (q)PCR amplification). GeneXpert cartridges are used to give a quick molecular readout for infectious diseases and can be deployed in the clinic.

The GeneXpert does this using a simple injection molded cartridge that costs <$20. The closest we’ve seen to this in sequencing is the iSeq 100 (a shocking unpopular sequencer). This also uses a self-contained injection molded cartridge with not fluidic coupling to the instrument.

Could we build a iSeq which performed not just the fluidic steps required for sequencing on instrument but also the sample and library prep?

I don’t see a huge technical barrier here. But so far vendors don’t seem to have found the market compelling enough to address. In the research space, it might be “kind of interesting” to be able to stick a raw sample into an instrument and get a whole genome out the other end. But the approach really shines in the diagnostic space. Here you’d be driving sequencing towards diagnostic applications. Along with cost, fixing automation issue could be the key to getting sequencing into the clinic.

My personal desire is to see sequencing ultimately displace qPCR. At first as a replacement for “high-plex” infectious disease panels currently performed using molecular testing (like the BioFire Film Array) but then hopefully making it’s way into wider applications.

This will require both vastly cheaper sequencing platforms but also “sample-to-answer” platforms using tightly integrated and cheap approaches such as injection molded cartridges.

This will hopefully lead to…

More Sequencing Based Diagnostics

It’s been both encourage and disappointing to see how (NGS) sequencing has been adopted for clinical applications. Encouraging because back at the when the first NGS sequencers launched the adoption in even advanced diagnostics was zero. There was no existing application that was a really good fit for NGS and it could just slide into…

Now Illumina report that clinical applications are ~50% of their market. A good chunk of this is NIPT, but the bulk is in oncology. Here NGS is being used to classify tumors and monitor for re-occurances.

But also disappointing because getting clinicians to adopt sequencing in other areas seems like an up-hill struggle. Coming from a tech background this has been hard for me to understand… but as I read more reports and personally interact clinicians the picture has become clearer…

Clinicians don’t view their jobs as finding out what’s wrong with a patient. They view it as improving patient outcomes.

Now… for the clinicians reading this this might sound confusingly obviously. But you have to understand the attitude of scientists and engineers. Primarily we want to understand systems, then if we can fix them. But diagnosing and understanding comes first and has value in it’s own right.

From a clinical perspective, the goal is to improve patient outcomes. As such many clinicians may view a diagnostic as unnecessary if it only increases our understanding of an individuals underlying condition and doesn’t alter treatment. And indeed if there’s only one viable treatment for the patients symptoms, why test at all?

While this is often the case with less common conditions (“flu-like symptoms”). It is also the cause for more serious conditions like intellectual disability for which in many cases, clear genetic causes exist. Many clinicians will be unwilling to test because while it may be possible to establish a genetic cause, treatments are not available.

However, my hope is that we’re starting to see a cultural change here, brought on in part by patient and clinician experiences during COVID19. As one clinician put it:

“Frankly, patient experience scores increase when patients leave our care and have a name for the thing that is bothering them even if there isn’t a targeted antibiotic or antimicrobial that has been identified, that would work against the pathogen. It still is comforting to patients and their families to have a name for the illness that they have.” - Emily Volk

The reality is that patients like to know what they have, even if an effective treatment isn’t available. One would hope that COVID19 has made the general population aware of the fact that you can perform diagnostic tests cheaply and effectively…

Broader adoption of sequencing is therefore unlikely to occur in a vacuum. Either we’ll require a broader array of therapeutics (so that identifying the underlying genetic cause or infectious disease does, in fact, alter patient outcomes). Or we need to see cultural shift, where it becomes unacceptable for patients to leave a care without at least a clear name for their condition.

That is to say… at some point in the future I hope it will be unacceptably weird to say “you got a bit of a cold” rather than, you have a “HRV-A2 CA90-0150” infection. Or to tell a child they have intellectual disability without identifying a clear genetic (or other) cause.

1. For some reason it has a combination of Ronald Reagan’s and Ayatollah Khomeini’s CG animated heads…