Cepheids and Sequencing

First and foremost, belated apologies for not providing regular updates to the Substack over the past few weeks. Life got increasingly busy with some fascinating conversations in relation to Reticula among other things. The sequencing world has been pretty busy too, with Genapsys filing for bankruptcy and Illumina being ordered to pay MGI $325M.

But rather than talking about any of that, today I’m going to write about Cepheid, a qPCR diagnostic company. I’m not inherently interested in qPCR, but it feels like Cepheid might indicate the direction we need to go in with sequencing to gain wider adoption, particularly in diagnostics.

Cepheid (or should that be Cepheids?) make molecular diagnostic (qPCR) machines, and have been around for 26 years. Their install base is 40000 instruments (about twice as many as Illumina). Annual revenue is ~$2B, with ~36 million tests shipped a year (compared to ~$3B for Illumina).

Critically, Cepheid has nothing like the market share Illumina has. They like have <10% of the molecular diagnostics market. But they do have a very simple sample-to-answer qPCR platform, which can be deployed in hospitals and clinics.

What would it take to build the Cepheid of Sequencing?

For many of Cepheid’s tests you simply bung the sample in a cartridge, load it in to the instrument and wait for the result (which takes ~45mins). We don’t seem to be anywhere near being able to do this in the sequencing world, where complex sample and library prep workflows are the norm.

Is there anything we can learn from Cepheid’s development of a sample-to-answer qPCR platform? Could we imagine a Cepheid-like instrument based around sequencing rather than qPCR?

Cepheid’s journey to an integrated sample-to-answer platform wasn’t direct. Their first product was essentially a qPCR machine with a fancy reaction tube:

Kind of neat, in that the design allows for better thermal transfer and therefore faster thermal cycling. But I would imagine not entirely game changing.

The instrument these tubes were used with is also neat. It’s composed of a number of independent modules each of which plugs into a backplane.

The Cepheid operator manuals even provide instruments for removing a replacing individual modules, which seems like it could potentially cut down in field support requirements:

Even this, relatively modest iteration of standalone qPCR instruments, seems a step beyond where we currently are with sequencing. But these instruments don’t allow processing raw samples. To enable a complete sample-to-answer system, Cepheid addressed this next, integrating sample prep into the platform.

They did this by building a fluidic cartridge which can process raw samples and also integrates their qPCR reaction tube:

Reagents can be preloaded, with no fluidic coupling to the instrument. The instrument interfaces with the cartridge via a reagent selection value and plunger (as described in this video).

The cartridge cost of goods, including reagents is ~$3 and kits are sold for $20+.

Current Sequencing Prep Solutions

In the sequencing world we’ve addressed automation around sample and library prep in a couple of ways. Firstly, there are pipette robots. In general these seem high maintenance, in many cases I imagine they require trained staff to develop bespoke solutions for each lab. But they are flexible, and generally open platforms. So users don’t get locked in to particular reagent suppliers. I imagine there’s also a relatively clean transfer from manual workflows toward automation, using the same lab supplies.

There are also instruments like the MagicPrep NGS. This more or less looks to me like a mini-pipette robot with pre-define reagent kits and workflows.

Digital fluidic platforms like the Illumina NeoPrep, ONT Voltrax, or Fluidigm Juno are available too. These generally seem pretty neat, quite versatile, but also expensive. With the Juno costing ~$15 per sample.

There doesn’t seem anything like the Cepheid cartridge. That is a cheap ($3) dedicated fluidic cartridge which does performs all necessary library and sample prep steps.

Why not?

Not a good fit for current sequencing platforms?

From what I can tell, there are no serious technical limitations to building a simple plastic injection molded device with preloaded reagents which would cover sample and library prep (but if you’d like to challenge me on this please get in touch (new@sgenomics.org)).

It would however naturally be limited to single workflows. For example, take a sample, extract RNA, reverse transcribe, add adapters, and generate prepped material for loading onto a iSeq or Miseq.

You wouldn’t be able to easily add sample multiplexing or otherwise modify workflows. And therefore you’d need sufficient demand to justify such fixed workflow devices.

Currently, I doubt this demand exists. Sequencing users either switch up workflows regularly or run fixed workflows on large batches of samples (where pipette robots probably work reasonably well).

Cepheid of Sequencing

The Cepheid of sequencing would therefore require a new generation of instruments, that would be addressing sample-to-answer applications in hospitals and clinics.

Like the Cepheid instruments it would be desirable to build modular platform that could scale with user requirements.

Instruments that fluidicly couple to the flowcell (like current Illumina sequencers) probably won’t work here. Everything should ideally be contained within the cartridge. This makes the system more robust, minimizes field support, and makes the platform easier to use and maintain.

One route is to use a Reticula-like approach integrating this with a Cepheid like molded plastic sample prep system. You can also imagine that it might work well with PacBio or nanopore detection platforms (the issue here is getting the COGS low enough).

But I think if sequencing is going to gain significant adoption in diagnostics, we likely need to radically re-think our approach to sequencing.

I’ve been talking to various groups about how we might go about this. Please feel free to reach out either in the comment section below, on twitter or via email (new@sgenomics.org) if you have any thoughts or suggestions.