Sept 15, 2005 | Five years ago, Nick Giannasi was studying the molecular phylogenetics of Asian pit vipers at the University of Bangor in North Wales. Perhaps sensing a lackluster market for population geneticists specializing in snakes, Giannasi gained experience in venture capital and investment banking, as well as spells at bioinformatics companies Doubletwist and Silicon Genetics, before joining Amersham. Following Amersham's acquisition in April 2004 by GE Medical Systems (now GE Healthcare), Giannasi was appointed Head of Informatics for GE Healthcare's Bio-Sciences division (GE Healthcare has about 40,000 employees and $15 billion).
| ||Nick Giannasi|
Aside from a major 2004 deal with GlaxoSmithKline (GSK), GE has been in "stealth mode" regarding its informatics capabilities. Giannasi dropped in to Bio-IT World to discuss how GE Healthcare Bio-Sciences is shedding light on data integration in discovery and clinical trials.
Q: How would you characterize the success of the Amersham acquisition?
A: It was a very natural fit. If you think of drug discovery, trials, and point-of-care as three steps, GE Medical Systems was at the point-of-care - hospital scanners, IT solutions, and so on. The Amersham portfolio was complementary at the point-of-care with the diagnostics - the juice that runs on the scanners - and the discovery systems business covers the first stage of pharma discovery.
If you can link those three [areas], there's not just revenue and cost savings synergies; there's the ability to provide more of a link to integrated discovery, trials, and down to point-of-care. That's the vision of our CEO, Sir William Castell - delivering personalized healthcare. By understanding discovery - data-generation technologies, instruments, consumables, IT solutions - and linking that into trial design and biomarkers, you'd be much more into targeted therapeutics. That will lead to more successful, efficacious therapies. That's the broader vision, I think, shared by a lot of pharma companies and consumers. But to get there, in a profitable way, is not trivial. GE acquiring Amersham is a very good mix of cultures.
How is GE Healthcare organized now?
It's split into two functional business units - GE Healthcare Technologies, headed by Joe Hogan, and GE Healthcare Bio-Sciences [10,000 employees, $4 billion revenue] under Peter Loscher. The informatics business (in Bio-Sciences) has a couple of focal points - we see life simply in terms of technology, classic IT stack, and then there's also vertical domain focus. The areas we're focusing on are principally enabling biomarker and lead discovery, and within that, particularly high-content screening. Another important area is translational medicine. Our role is to bridge the Amersham business and the medical side. Our technology focus is middleware and data integration, not just within discovery but leading into clinical and preclinical. This is an increasingly active role in our group.
What's your relationship with GSK?
We've been working with GSK for the last 12 months on the next generation of DiscoveryHub, our data integration middleware, and that's gone very well. We see that as a really good platform to go out to other pharma companies and biotechs because the [current] market is very fragmented. We saw companies like Spotfire doing well on the top, various players on the bottom, screening with IDBS, [Elsevier] MDL in chemistry, various LIMS [laboratory information management system] players. But we didn't see anyone doing the middle bit particularly well.
Instead of reinventing wheels, we're trying to leverage technology. But technology is only half of it - there are human issues. Trying to change people's way of working is not trivial, especially in a multinational pharma going through mergers. So why not try to improve the way people work, rather than trying to radically change the way they work, through technology?
What exactly do you mean by "middleware"? Is it glue that integrates third-party tools?
It's one of those broad terms - ours is principally a federated query-processing engine. If you took something like Oracle and split it in half, the top half is query processing, the bottom half is physical storage. Remove the bottom half, it sounds like IBM DiscoveryLink. There are similarities [with DiscoveryLink], but here's where it's different. In other industries, the actual data that sit below are not that complex. However, in life sciences, there's a lot of data that is very nested, very hierarchical. Our data-integration middleware is designed to deal with those data from the ground up. We don't see it as creating a new market for federated middleware that's going to change the world on its own in pharma discovery by any stretch, but something complementary to existing warehousing and more classic integrated approaches. It's not just a pharma trend; we see that in other industries - Sybase acquiring Avaki is a good example.
What proportion of your revenue stream is product versus consulting services?
A typical engagement is 75 percent license and 25 percent service, which is manageable from both sides. DiscoveryHub comes with a set of specific wrappers or connectors to data sources. It also comes with generic ones for database management systems (XML, etc.). To add a new source is very small in terms of time - a couple of hours or days depending on complexity.
Who buys your products and services?
Historically, people have pigeonholed the IT and science guys, and alienated one or the other group. We see it as a virtuous circle with three areas - the scientist, the IT, and the business user, who could be a scientist or IT, but can also be a third party. Sometimes in pharma, we see the IT and informatics departments as pioneers; sometimes it's the scientists or the senior business leader. We're trying to understand how the two interact. What we have at GE Healthcare is discovery and clinical trials, so we can help scientists get access to more data more quickly to make a more informed decision. That's our basic premise.
How does DiscoveryHub middleware compare to IBM's DiscoveryLink, Oracle, and other bio-IT software?
DiscoveryLink and DiscoveryHub are the only specific, query-based, federated engines aimed at life sciences. That's our genuine view. How they got there is completely different - ours was built for that. We see InforSense as very complementary [to DiscoveryHub]. We can pop into things like InforSense and Spotfire. There's no point being competitive with Oracle - it's a massive installed base, and there's a huge number of trained Oracle DBAs. What can you do that Oracle doesn't quite do as well as is needed? That's where we come in.
How does your model interact with chemistry software?
Companies like MDL, Accelrys, Tripos have great technology, back-end data management, explicitly good for chemistry. We can add in extra elements, linking it to more of the biology and clinical space. We would never be replacing IDBS's ActivityBase; we help people get more out of their ActivityBase data and investment. LION's SRS is just a data source to us.
What is the breadth of data sources you handle?
It's changing. Historically, we looked at more complex, nested types of biological data. Over the past couple of years, we're seeing requirements in what we'd thought were more stable areas such as chemistry. Now, it's medical imaging data. A couple of years ago, clinical trials and translational medicine was about 5 percent of our business. Now it's 50 percent and growing. A lot of the trial data is image data, but it hasn't historically been put in context with other data.
Are you only interested in big pharma companies? How do you look at the customer universe?
We have a very simple pricing and license model. There's a minimum of 10 users, you can add in any value, you can store it on as many servers as you like. It's a perpetual model.
We don't go into a pharma and say, "To make your pipeline impact and change, you need to have integrated enterprise and we will spend five years doing this at a cost of X" - they'll just laugh you out the door. The vision is, take any pharma company and slice it into discovery and trials, under that biology, chemistry, animals, humans, under that enabling technologies - genes, proteins, etc. It would be fantastic if all of this was integrated across silos, geographic locations, and therapeutic areas, but let's not go there yet. You can put hotspots where you can make more informed decisions. Lets take one of those, prioritize what's important, and we'll work on one of those, compare the value, and that could give you more confidence to bite off another couple. Our success is you guys being successful, showing you this "before and after." Then you're going to walk to an enterprise deployment over a number of years, because you're going to be around for a long time, and we're planning to be around for a long time. It's from selling middleware in year one to an enterprise environment after five years.
What does a DiscoveryHub installation cost?
A typical installation is $75,000 for the [perpetual] licenses and $25,000 for services (within a 5 to 15 percent range). There's also an optional 18 percent ongoing annual support and maintenance charge. DiscoveryHub (5.4) is 9.5 years old; it's reasonably mature. The next major release will probably be 6.0 - it has some significant improvements.
DiscoveryHub is the main focus of my informatics group, but we do more. Pharma companies traditionally feel that, when one vendor can't supply everything, they need to get an Ocean's Eleven together of different technologies. But who is accountable to deliver that? They've felt a little bit let down. So that's another aspect we do - if we go in with IDBS, we'll take a portion of that revenue. Another recent deal is with Gene Logic, which recently brought out an ASP variation targeted exclusively at academics at a much lower price ($3,500 for an annual subscription). My group will be selling this from now on.
Is business good?
Business is improving, but it's been challenging over the past 18 months for a number of reasons. There's been over-promising and under-delivering from the vendor community. Middleware has probably been as bad a culprit as any of these areas; there's a certain amount of market skepticism. So we've said, let's collaborate with firms like GSK, actually build some value, get some solid metrics, and then we'll tell people and let the market decide. A back-to-basics approach. I think things are very exciting right now - our RFPs have become much more sophisticated. The R&D leadership is more in tune with the IT/informatics.
What global events could most impact your business positively?
Pharma companies trying to move away from a blockbuster model and submit biomarker data, though they're being encouraged to do so by the FDA. It's a tough one. If the biomarker became an accepted way of shortening trials, that would significantly shift the translational medicine focus.
What could be the most disruptive event?
From a technology perspective, if the acceptance of federation just vanished, and data had to be put in one place. We can physically store data, but it's not where we want to go. From a people perspective, we've seen acquisitions by pharma [have] an impact on cultural silos. You still have a scientific culture where sharing data and knowledge is a tough one. If that doesn't become accepted, that would make our impact pretty insignificant.