General Manager, QIAGEN Waltham
This executive has a background in nucleic acid amplification, next-generation DNA sequencing, microfluidics, genetics, molecular biology and product development, with experience in innovative technology development from the initial start-up to product launch stage. He is currently the SVP R&D at IsoPlexis, and previously held senior roles 4Catalyzer and Quantum-Si and Qiagen among others. At Qiagen he was Head of the global Next Generation Sequencing (NGS) development team, responsible for management of the technology development driving QIAGEN’s NGS GeneReader system.Read moreView Profile Page
I'm a generalist, but I have a strong science background. In part, for me, I heard about CRISPR in 2013, and the more I learned about it, the more it became clear to me that if we were able to turn this into therapeutics, it would be revolutionary. I delved deep into biotech and therapeutics, which led me to believe that genomics would open up the next huge revolution in medicine. We're now at the forefront of the next big area of medicine, and it all points back to Illumina for me, in terms of having the biggest market share and driving pricing down to a level where you can build applications that will then pull demand. We're at the cusp right now where the economics of NGS is a no-brainer, where for $200 or $300, everybody will have a genomic test every five years or something. I'd love to hear your feedback because you've been in the industry for a while, and I'd love to hear about the history a bit to understand what proteomics may do and how the industry may evolve.
I want to learn about the evolution of the NGS industry. I know that Qiagen had their own gene reader, then they abandoned that and went with Illumina, and I want to understand why you think that is. SomaLogic talked about an instrument they use, but they’re working with some key technology leaders to transition to their reader. I assume that's Illumina's NovaSeq, probably, but I might be wrong. I wonder how proteomics can work on an NGS reader. I don't understand how that works. It may just be that you have a different flow cell array that you’re reading off of, and that's all it takes.
Really cool stuff. It’s too bad we’re not co-located because this would be a fantastic conversation over several beers.
Let’s start with some history and why Illumina kind of won. What is their positioning today relative to some competitive things like PacBio, long reads, Nanopore. I'm more interested in the history and how we got to here. And if we have time, where proteomics is today and NGS, and what you think the industry evolution might be like.
So right out of post-doc, I got hired at 454, which was the first to have a commercial sequencer; Solexa/Illumina back then had vaporware. They were always talking about this thing that was going to come out. We didn't have anything, so we came out with this first real sequencer. Back at the time when no one knew what to do with it, there was this whole question of what's this even used for.
That was just after the Human Genome Project was completed, right? That came out of that.
It came out an interest in it actually spun out of Purigen, which was a weird kind of sequencing for hire, finding genes of interest for companies. They were ahead of their time; they had banks and banks of huge slab gel sequencer type stuff. A caveat for anything I say, if you look at the technologies that I've picked, I've been famously wrong. Illumina was coming along with their 25-base reads, and we had 100-base reads, and we were like, this is cool. There was the issue in the homopolymers, but we had longer reads, and Illumina just completely ate 454's lunch for two reasons.
This was before or after Solexa?
After they merged. They merged early on. Probably in 2004 or 2005 Roche bought 454, and there was no interest in downward pressure on the costs. It was always more expensive. And that's a very hard position to negotiate from, to argue that our bases are worth more, and Illumina was producing tons of data. It was pretty crappy quality back in the day because it was cheap and could produce a lot of data. There was a lot of interest in the academic world on pitching in and writing better algorithms. They’ve done miraculous things with enzyme evolution to make the system to where they're at; you could get 300 bases. Now, 454 was doing 800 bases at the end before they went under, but it was just kind of the most beautifully advanced buggy whip. They just weren't cost-effective. We can discuss long reads versus short reads in a bit, but it was a fast sequencer that gave long reads. So you could argue that with a different incentive to the market, maybe they would still be in business, in a niche business, because there are certain applications where you want that kind of length. It's nice for HLA typing, that kind of thing. Now you're clocking something relevant.
So then came Ion Torrent, which was, in many ways, the same cast of characters. We all bailed out of 454, went and did something else for a couple of years, and then Jonathan called everybody and got the band back together, and we started Ion Torrent.
Did that come out of MIT?
Yeah, the MIT review is kind to us in running the articles. There was certainly that ISFETs and stuff, and probably there was a heavy MIT component in that. But putting the sequencing together was completely homegrown and a lot more precarious than one might imagine. We got it to work, but there was a long time where we were considering what else we were going to do because it just wasn't working.
What do you mean by that? What wasn’t working?
It was interesting. The whole thing effectively used the ISFETs as a whole bunch of tiny pH meters, so when you incorporate a nucleotide into your growing strand, as you're sequencing, you've got your template out there. The polymerase is sitting on it, and you're feeding nucleotides, and it starts incorporating nucleotides to copy the DNA. Two things happen. Once you release a phosphate, the PPI comes off, and that's what 454 was catching, this luciferase, which would then cause a glow of light. Pyrosequencing does that as well. But what also happens is a hydrogen ion is lost. So Jonathan had stopped me a long time ago and said, if you worked in a bufferless system, could you chart that pH drop as the thing becomes more acidic? And I said, no, of course not, that's just silly; you need to buffer everything. He said, well, why do you need to buffer everything? I said, every polymerase is sold in a buffer, so of course, you need these buffers. I mean, why else would they sell these buffers? It turns out many of these polymerases work just fine in no buffers whatsoever.
What do you mean by a buffer? Is that a solution?
To maintain the pH, so they're most happy and pH balanced, but that's often because you've got them sitting in a tube for a long time and you're thermocycling them. But if you're constantly flushing new material through, they don't care. They're quite happy. So in this ISFET chamber, when the proper base comes and is incorporated, you produce hydrogen ions that drive your pH down in a method that's completely correlated to the number of those incorporated nucleotides. If it's only a single A, you get a 1x pH shift. If it's two As, you get 2x; it's not perfect, but it's pretty close. So suddenly, we could use native nucleotides. We didn't use fluorescents. The data is coming out from the CMOS chip itself, which is just reporting voltages, so there's no image acquisition. It was and is fast, and there the company was acquired by Life Technologies, and they went in two directions. They had a small kind of desktop sequencer. I can't remember how many wells that had, maybe 1.2 million or something. Not a ton by today's standard because you can't fill every well; it's Poisson distributions and whatnot.
But then they went bigger and bigger and bigger because there was this kind of chasing Illumina, so scaling up these chips larger and larger. But as the chips get larger, you also typically try to shrink the size of the well. Noise remains constant, and your signals get lower; there were some real challenges. I don't know those challenges very well because I stepped out of the company right after we got acquired by Life Tech. I had to do two years there for contractual stuff, and then I hopped out. I am just really interested in the big company thing. But looking back, I'm seeing how that system is being used. I think there was an opportunity, exactly what you're saying, that people will be sequenced regularly, and it's probably not going to be a whole-genome sequence. It's probably going to be a targeted panel of things.
The cool thing at my last job at Ion was I developed the AmpliSeq process. That was my baby, and suddenly you could amplify 40,000 things in the same tube. A cool project. Now Thermo has licensed AmpliSeq to Illumina, and Illumina runs AmpliSeq on their systems as well. There's this market, I feel, for a high-speed smaller, targeted sequencer that gives decent read lengths. Again, we're talking with Ion, and I don't keep track anymore, but they are probably in the 200 to 400 base pair range.
But there's no denying the fact that Illumina is the 800-pound gorilla. They've kept pushing the needle. They've kept getting larger and larger systems, but then they also went smaller, so they had that ability to cover the bases. They tried a sample prep system which did not work, and the NeoPrep was a great concept, terrible execution. They’ve covered what we’ll call the short-read area very broadly, and to the point that runs take them a long time, they take a couple of days when you're running the whole thing, but you get so much data. Frankly, it's hard to envision an emergency room doctor saying I need this genome sequence, stat, and they run off. It seems like you have that time available. But there are niches that are being exploited, and I think one of them is Ion, who is still a contender. People seem to like the speed of that turnaround, and it allows you to do more single samples without having to pool. Illumina is cost-effective if you're using that chip to capacity. The minute you start saying I'm just going to run a single lane, suddenly that single lane starts getting relatively pricey.