CLEVELAND--Manuel Glynias has some educated hunches about the direction the bioinformatics field will be taking. "One nice thing about my job is I get to talk to lots of pharma company VPs all the time," said Glynias, CEO of bioinformatics firm NetGenics. "Sometimes they're all saying the same thing, other times you can hear some really interesting differences."
It's been nearly three years since Glynias and his partners put together the business plan that helped them bring in the initial $1.5 million in financing that got NetGenics off the ground. After using the cash to build its cornerstone product, a bioinformatics software integration package called Synergy, the company won a few big customers and succeeded in raising another $25 million. Now, Glynias is steering the firm to respond to bioinformatics needs he sees coming in the post-genomic era.
Glynias spoke recently with BioInform about his vision of the future of bioinformatics in drug discovery and about his strategies for prospering as competition in the field intensifies. The first part of this interview appeared in the January 18 issue of BioInform.
BioInform: What are you hearing from pharmaceutical executives about their bioinformatics needs?
Glynias: In the same way all the pharma companies went through this wrenching decision about whether they should spend money on genomics, the next big area now is pharmacogenomics. There are some companies like Glaxo Wellcome that are already spending money, and there are others that are really holding back. Some are wondering, could we achieve the same sort of thing a lot more quickly using some other technology?
One of the things I'm hearing lately is, maybe we can do this with expression analysis. Can we use Affymetrix chips, look at patient populations, and look at responding populations and nonresponding populations as predictors of whether my drug is going to work?
It's the same kind of predictor they want to have using the single-nucleotide polymorphism (SNP) maps. But we could have this today; we don't have to wait a few years until the SNP maps are done. It's a reasonably easy task, the technology exists today. Nothing new needs to be invented. Companies are starting to differentiate themselves, at least a little bit now, as far as the approach they want to take post-genomics.
BioInform: What will be the role of the bioinformatics provider in this next wave?
Glynias: What we're hearing is that there is a tremendous need for software for understanding expression analysis, more than there was for genomics. Even with the simplest chip you're looking at 20,000 genes at the same time. There's not much good software yet to be able to help a scientist take this set of 20,000 and get an understanding of the patterns. A number of companies, including us, are working on such software.
What we're trying to do is make tools that make it easier for biologists to take this brand- new data and relate it to data that they have some biological intuition about. What's cool about this industry is that a year ago there was essentially no need for that. You would have had to be more visionary than I was, I'm afraid, to guess a year ago that that was the thing that people needed next.
BioInform: How far along are you on developing such tools?
Glynias: We started working on Affymetrix data almost a year ago, but it wasn't clear to us how fast it was going to take off. Now everybody wants it and they want it yesterday. So we're working hard to get the little tools we have expanded to be more useful to people.
BioInform: Will these tools become part of your Synergy package?
Glynias: Part of the idea of Synergy is that we create a tailored version for every customer. American Home Products' interests are in one direction, other companies' are in different directions. We build the pieces they want in a component style so we can reuse them for other customers. Just think of Synergy as the IT infrastructure that tools like that would hang on.
BioInform: So Synergy will accommodate other bioinformatics vendors' components too?
NetGenics: This whole idea of components and the Object Management Group initiative is going to make the pie a lot bigger. It's going to convince pharmaceutical companies that there are lots of good tools out there and they could buy the very best tool from every vendor because we can hook them all together and get a better system, not a confusing system.
We're already having talks with third party companies and they're very interested in talking with us about ways to make our software interoperable. A lot of people really aren't all that competitive when you look at it. In other words, if you look at the strengths of these various companies, they are often quite different. If you're willing to write the thing you're best at, create the tools you're best at, and, in our case, provide the services you're best at, lots of people can be successful if companies can buy the pieces they find most useful and they all hook together.
BioInform: Who are your competitors?
Glynias: Pangea, I guess. The difference between our model and Pangea's is that Pangea is quite happy to sell you a piece of software. We don't do that, but we're competitors in the sense that we're competing to get the same money out of the pharma companies.
The same could be said for Incyte and Molecular Applications Group. Molecular Simulations also has some software in this area, as does Oxford Molecular. None of them have taken our approach, and we think our approach is unique enough that we can actually halfway coexist with all of those. With the exception of Pangea, we really are talking to most of the other vendors.
BioInform: How, exactly, is your approach different from the others?
Glynias: What they want to do is push a new algorithm or a new database. We don't do that. What we do is figure out how to integrate all those things together so a scientist can ask a more interesting question.
Of course, the rest of what we do is to let you remember what the answer to that question you just asked was and share it with your colleagues. Our system remembers every change ever made and the name and the date and the person who made the change. Those kinds of elements most other bioinformatics companies have ignored because they're interested in doing other things.
BioInform: Will you describe how you are doing this?
Glynias: A lot of companies have one central server containing all the data, and everyone using that across the company. The major problem is bandwidth. So, what people then do is replicate that server to other places in the company. They'll make an exact copy of the server at different sites and the changes of data get copied to the other servers once an hour or once a day.
The trouble with that is there's no bandwidth problem, but there is a communications problem. If I'm in New Jersey using the server, I don't see the data on the Switzerland server. I certainly don't get any minute-by-minute contact with that thing. Maybe every night it gets justified, but there's certainly no real connection between them.
We came up with the solution, which is to take the best parts of both those and combine them. Servers at three locations are continuously in real-time contact with each other. They're not making copies of the data, they're all owning certain parts of the data and making them visible to the whole system.
Let's say you have a giant computer that's the perfect thing for running Blast searches. You spent $1 million and bought the best computer you can buy. You can put that at one of your sites, let's say New Jersey, and the people in California and in Japan use it without even knowing they're using it. When a request for that particular thing occurs, the server knows that the most effective place to do that search is actually in New Jersey, sends it there and brings the result back, stores it locally, shows it to the user. No one's ever done this before, at least not in bioinformatics.
BioInform: This sounds more like general IT than bioinformatics. Why wouldn't pharmaceutical companies have their own internal information systems staff handle this?
Glynias: Well, it's both. The system that we've built does general IT, but it does it in the context of bioinformatics. It's a system that has what we are calling nondomain-specific components, but on top of it and integrally related with them are the ones that are effectively specific.
Objects have behavior and they also have inheritance. So we have an Incyte DNA object which inherits all the properties of a regular DNA object but also has peculiar things that make it an Incyte object. We built the entire system from the ground up on this kind of technology. And using this sort of technology gives these companies a competitive advantage over technologies that their in-house IT groups can put together.
If they were to try to build this internally it would cost them as much as it cost us to build it internally. We've raised $25 million in venture capital so far. It would be hard for them, even for $10 million, to build an IT system like this. Why should they when they can license it from us for some small number of millions of dollars and have it be our job to install it, maintain it, and train their people on it?
That's why we've gone to this outsourcing model, which I think is still fairly unique among bioinformatics companies. We see that pharmas are increasingly trying to concentrate on just what they do best--develop and market drugs. We saw the biotech investment boom and the CRO boom and said look, it's clear that a strategy these companies are taking is to outsource everything that's not core. It's certainly not core to build an information system like this.
BioInform: What is your relationship with Incyte?
Glynias: They're one of our investors, and they gave us the license to be able to understand how their system works well enough that we can integrate it into Synergy.
An interesting problem with Incyte customers is that of 23 pharma companies, all have exactly the same data, and they're using exactly the same tools to analyze it. My guess is they're getting exactly the same answers. Presumably they're all working on the same target. So from our point of view, a valuable thing is an upgrade to the Synergy product that has hooks to Incyte in it. Distributed Synergy allows us to have eventually seven servers all talking to each other, linked so a user thinks it's one giant computer.
A customer of both NetGenics and Incyte can do two things: either work inside of Synergy and pull data out of Incyte's databases, or work inside of Incyte's LifeSeq, do some preliminary analysis there, and cut it down to 10, 15, or 20 sequences that they're interested in. We've modified Incyte to add a button that says export to Synergy.
You can also do database searches against the Incyte database inside of Synergy. All of a sudden, a pharma company has access to both sets of tools and can go back and forth between them to figure out what they want to do. NetGenics is the first third-party company to be able to do that with Incyte.
The other part that's exciting is we model the Incyte database inside of Synergy so you can annotate Incyte data for the first time.
BioInform: Would you set up the same sort of service for a customer who was a subscriber to a database other than Incyte's?
Glynias: Certainly. Presumably one day we'll be asked to do something for American Home, which is a customer of Millennium's. Celera says that someday they will have a lot of data, too.
One of the advantages of being independent is that we can go talk to Millennium, or Celera, or Gene Logic. There are a lot of companies that are looking at Gene Logic's technology, Affymetrix's technology, and Incyte's Synteni technology all at the same time. Some of them are planning to use all three of them. They would love a system where they could, for example, integrate expression technology. We go to a pharma company and no matter who it is they deal with, we can go talk to all of them.
BioInform: What else is on the horizon?
Glynias: We have a chemistry module to Synergy that we've been working on. What we'd like to do is partner with cheminformatics companies to see if we can build tools that do a really good job of integrating biological questions with chemical questions.
As far as I know, there are no systems that let you ask a question where some of the databases being queried are chemistry and some are biology ones. Increasingly those are the kinds of questions that scientists in drug discovery want to be able to ask.
Cheminformatics is a well-established industry. We're not interested in trying to be in that market. What we want to do is try to create the interfaces between the chemistry and the biology so that we can go to any drug company and say look, if you're using MDL or MSI or Tripos, we can work with any of those vendors so that the data you have in their databases can be integrated with the biological data that you're keeping.