Last week, OVP Venture Partners, a Kirkland, Wash.-based venture capital firm, teamed with Amgen Ventures to add $11.8 million to the pool of funding available to Accelerator, a biotechnology incubator affiliated with the Institute for Systems Biology. OVP was an early investor in Rosetta Inpharmatics, and more recently co-led a $4.3 million financing round in NanoString, a Seattle startup commercializing single-molecule bar-coding technology developed at ISB. OVP cited computational biology and bioinformatics — specifically “advancements in modeling, analyzing, and storing vast quantities of data to aid laboratories involved in genomic and proteomics research” — as key investment areas that it will target under the Accelerator partnership. BioInform caught up with Chad Waite, general partner at OVP, to learn a bit about what types of bioinformatics startups are catching the eye of VC firms.
It’s great to see investors looking at the bioinformatics market again. How much of OVP’s life science investment will be in software startups as opposed to other life science tool platforms?
I will make a couple of distinctions, and they are probably subtle ones, but this is how I view it, at least: I am not interested in spending a lot of time and money investing in what I would call single-function instruments. Why in the world does anyone in a venture firm need to develop a better mass spec? You’ve got plenty of people out there spending a lot of money with big research budgets that are doing pretty good stuff. So to me, that doesn’t look like a good venture opportunity.
I think one of the things venture folks are guilty of — and I don’t know whether this is good or bad — is that we think in paradigms. And I have two partners that were instrumental in launching the whole e-CAD [computer-aided design] business. So if you look back to the early 80s, you had what I think to be a very similar set of circumstances in electronic CAD that you do today in biology.
In what way?
Before [e-CAD firms] Mentor Graphics and Cadence and several other companies, you didn’t have the people that brought together the computer hardware and the sophisticated software to design and simulate electronic circuits. And if you look at the increase in computational speed and complexity since 1983, it’s obviously many orders of magnitude, and the kinds of problems that you have in modeling pathways or modeling cells or modeling protein interactions obviously couldn’t be solved in computers ten years ago. So what I think we have today is new mathematical approaches, new software architectures, and new computer architectures that allow you to model extremely highly complex problems.
So I think the point you made, of why are venture people interested in this, I think mainly because venture people previously made investments in this area before the technology was ready to do what needs to be done. I think it was just too early.
Now I think that some of the opportunities are software platforms. If you look at proteomics, for instance, you’ve got three or four major different classes of instrumentation performing experiments, and at least in a mass spec run, one experiment generates upwards of two to three gigabytes of data. And people have largely been managing that either in internally generated open source software, or Excel. And, just one example, one program at ISB generates about 120 gigabytes of mass spec data a day. You can’t manage that in Excel.
So my theory is — and we’re out testing it, and we’re going to make some investments while we test it — is that there’s been just a sea change in the amount of digital information generated in these kinds of experiments. So what you have to do is develop commercially released, documented, powerful software to manage the data, group the data, analyze the data, and communicate the data.
So I just think we’re at sort of an inflection point where there are going to be some very interesting opportunities.
For us, it’s not going to be ten deals a year; it’s going to be maybe a couple, and some of them are going to be inside Accelerator and some will be outside Accelerator. Nanostring was outside of Accelerator, even though it came out of ISB, because [it was] further down the pipeline, if you will, in corporate development than an incubator would typically involve itself in a project. … So the first area I’m really looking at is software platforms. It’s not announced yet, and it’s not done, but we’re about to make an investment in a software platform.
Is that out of ISB?
No. this one has customers, and it has product. It’s got a CEO and it’s got 15, 16 people in the company. So it’s early stage — the product’s not totally finished — but there were seven or eight customers that we were able to talk to. Beta customers. It’s in the proteomics area.
I thought Accelerator was linked specifically to technologies developed at ISB, but it sounds like you’re casting your net a bit wider than that.
Actually, the two projects that Accelerator has started already [VieVax and VLST] didn’t come out of ISB. And there is one we’re working on right now that will come out of ISB, but that’s several months away before that one’s put together.
But Accelerator was meant not only for ISB-related projects, but also projects that find their way to [ISB founder and president] Lee Hood because he’s such a magnet. So it’s really a way to capture some of the intellectual property generated by Lee at ISB, as well as capturing some of the intellectual property that comes to Lee because of who he is.
Looking back at earlier investments in the bioinformatics market, it seems like VCs who came from the IT side were disappointed because the potential market for bioinformatics is so much smaller than it is for general-purpose IT, and the ones from the biotech side were let down because there wasn’t the potential for drug revenues that there might be for other biotech companies. Are firms like OVP more realistic about the market dynamics for bioinformatics software now?
I come at it from having invested in both sides. And about 30 percent of OVP’s activity for the first 15 to 16 years of its life was investing in biotech — not all of our activity, but 30 percent. So companies that I was involved in starting in Seattle were CellPro, Corixa, Seattle Genetics, Rosetta, and a number of others, so I come at it from understanding the biotech side, the drug development process, how you build a company in that area. Plus, the other 70 percent of our activities has been in computer software and infrastructure. So I have my feet firmly in both areas, and I really think that a greater portion of the IT budgets in major pharmaceutical companies are going to be spent on discovery tools and platforms over the next 10 years. So the perspective from the IT-only people was that the markets weren’t big enough, but I think that’s going to change.
At Rosetta, when we were starting the company, we used to sit and look at each other and laugh, saying what we’re really doing with all this expression array software and analytics, what we’re really trying to teach guys at the large pharmaceutical companies is a different way to discover drugs. Now, if we went out with that message as our marketing message, they wouldn’t even have let us in the door. But what we did is we really backed into it over a number of years because what major research company wouldn’t use expression arrays now? And that wasn’t the case when we started Rosetta.
And so today, what does Affymetrix ship? $600 million a year in arrays or something like that? So it’s a major piece of everyone’s toolbox today. And I think that proteomics is bigger, broader, it’s coming on faster, and it generates more data. So my theory is there is a lot of opportunity to build a couple of very interesting companies in this area. Not 50, but a couple.
You mentioned the biosimulation area. What do you see as the timeline for commercializing those types of software platforms compared to what you’re seeing in proteomics?
I’ve got a company in that area that I’m working on, too. I think that’s obviously a much further out issue, but there are some rather narrowly appreciated fields in mathematics where you can build algorithms and software architectures around which you can start to do dynamical complex systems modeling. And of course, it’s still highly speculative, but I think there are some research projects … that are going to generate some very, very interesting models. But I think as far as making a contribution in the whole pharmaceutical food chain, it’s down the road. But that means you should start thinking about that stuff now. The market’s not here, but that’s the question that we always pose to ourselves, and I think a lot of other people are — folks that run large divisions in large pharmaceutical companies or biotech companies, they haven’t seen proof yet that by the use of these tools, you can actually decrease costs and lower the time horizon that it takes to develop a drug. Once I think someone does it, there will be hopefully somewhat of a stampede.
The 777 was the first airplane totally designed on a computer, but are they going to go backwards? Are they going to go back to building balsa wood models? I think that once you sort of get over the top, you create a totally new market, which happened with arrays. But I think that [the biosimulation] market, in our looking at it, that’s another five to seven years away. But that means we have to start thinking about it now.
What are some other emerging opportunities you see in the computational biology market that look promising, either short term or long term?
I think there is an interesting opportunity in creating … data portability. I think that in this market, there’s no XML yet. It’s that XML for this market is not XML output for a business system. So I think there’s some interesting work being done around creating a biological XML.
Several of them, actually, right?
Or several of them. They’re out there, but none of them are being adopted yet. And that could be due to a number of reasons. In the proteomics area, every instrument outputs its data in a different format. And not only is it every different instrument between, say, chromatography and spectroscopy, it’s also different brands of mass specs output their data differently. So you have no common model yet by which people are able to express data sets.
And, like you said, there’s been a lot that have been put out there, but none of them have been jumped on. And a lot of them have been put out there by academic institutions. So that means that they’re not properly documented, they’re not maintained, they’re not upgraded. I think that it’s like an Adobe coming out with Postscript. You need someone that’s commercial to take the bull by the horns and make it a standard.
Do you see any candidates who might be able to accomplish that?
I’ve seen one, but it’s an academic one, and I’m not sure where that’s going to go.
Is there anything else that you’ve seen in scouting out this sector so far that you think is promising?
I’m literally in the middle of five projects right now, and they’re all at various stages of being baked. But it’s everything from some proprietary IP around biomarkers to software platforms to simulation technologies, to some very interesting enzyme chemistry. So it’s sort of all over the map, and none of them are drug target-related single-molecule companies. They’re all building blocks.
It’s nice to see a VC supporting tool companies.
Well, the crystal ball is imperfect, obviously. But if you’ve been doing this as long as some of us have, and you’re not an optimist, it’s probably time to find something new for a living.