OXFORD--In part 2 of our exclusive interview with Tony Marchington, the Oxford Molecular Group CEO discusses the move toward platform independence in bioinformatics software, his company's approach to collaboration in software development, the trend toward creating software for less expert users, and his strategy for finding the best bioinformatics talent. Founded in 1989, Oxford Molecular recently completed the acquisition of Genetics Computer Group Inc. (GCG), the developer of one of the most commonly used software packages in bioinformatics, and of the software business of MLR Automation, whose software is used in managing automated high-throughput screening data. The company's other products include MacVector, GeneWorks, GENESEQ, TOPKAT, CAChe, Cameleon, and RS3 Discovery.
BioInform: Would it be a fair assumption that in just a few years, all the major bioinformatics software will be platform-independent, or will there always be certain proprietary applications?
Marchington: It's certainly going that way. Without making any fancy estimates, just look at the way it's trended over the last five years. There are still some peculiarities around that need to be addressed, and still there are differences between how these particular Unix boxes perform, and a lot of it depends upon the hardware suppliers themselves. Silicon Graphics is obviously putting a big effort into data mining, and there are still some specialties that Digital offers. You've still got some specialist hardware boxes around for genome analysis and stuff, there are still some oddities, but my own guess is that this stuff will largely go more and more platform-independent. As the Java and the other standards become totally platform-independent there's going to be a network in place. There might be some hardware around that seems better than the other pieces of hardware and you might get an intelligence in there that actually says I'm better going and doing this on the IBM or on the Silicon Graphics server. But in terms of the user, the user will want a very simple PC-type interface where everything just looks the same as his Microsoft Office and he can just pull this stuff down and use it and transfer it straight over to the documentation. It will be as simple as that.
BioInform: With Diva, you developed a product tailored for one of your large customers that was later released generically. Is that type of collaboration a trend in bioinformatics software development?
Marchington: There aren't two versions of Diva, there's just one generic version. As we go around the industry, many companies believe they've got special requirements, but in actual fact, by and large, 99 percent of the requirements are going to be the same. So if the industry develops specialty products for Glaxo Wellcome or SmithKline Beecham, that would supply 99 percent of everybody else's requirements, that's our observation.
We know that the way not to develop software, it doesn't matter which area you're in, is to take a lot of clever engineers and stick them in a darkened room in Oxford or St. Louis and say, "build me one of these." Because the complexity of what a customer needs cannot be attained in that way. The only way to get the solution is actually to prototype products at the heart of the customer's research establishment. What we've observed with the companies we're now working with, and it seems to be common throughout, is that we initially go in and say we're building a new product for you, what's the big problem you'd like to solve and why is it a problem? How would you like us to solve it? Then we'll come out with an initial spec. We go away and we prototype that spec. Then we take it back and say, right, we've prototyped it exactly as you've said and here it is. And they say oh, we don't want that, we want this. So you go and you've got your next generation. After four or five of those iterations of prototype and test and reprototype and retest, you actually get something where they sit down and say, yeah, we want that, we can't have it fast enough. Then the rush is on to engineer it, which is where the object-oriented and modular design problems come in. And what was very, very clear--both with Omiga and, on the small molecule side, with Diva--was that although they were engineered largely with one company in mind, every other company that's seen them is saying yes, it's a very high-quality product and certainly headed in the direction of how we see an ultimate long-term solution.
BioInform: There also seems to be a trend toward software that requires a lot less computer expertise on the part of the user.
Marchington: Yes, it's essential. The market for information technology in pharmaceuticals is following the same trend as information technology in many other industries. If you look at aerospace or engineering or chip design or even financial services, before information technology came along there would be a few experts. For example, if you wanted a Bloomberg report running in a financial institution in New York, and it was five years ago, you would go to an expert who sat in his own little area, and you'd say I'd like this report and he'd take down what you were looking for, and he would have this very complicated task of putting in the specifications of the report and making sure it was all hunky-dory, and sending it off. Then the report would be printed and you'd get it in the mail a few days later. That's sort of the early-adopter stage of information technology, where an expert used the machine and nonexperts go to the expert. And that market is always characterized by certain things; one is, it's always very small, because an organization of two or three thousand can get away with three or four experts, and that's the number of systems they have, and those first systems can be quite expensive.
But as the market starts to mature, and this is the same for PC's and office software and spreadsheet use and every aspect of IT, the move is toward nonexperts who don't want DOS, they want Windows. They're not bothered about getting an a:\ , what they want to do is just click on a window and get a list of stuff they need, then click on something else. They don't want to know how the computer's doing it, what the file format is, they just want to get a very simple presentation of what's available so they can make a simple choice and get the result. So what's evolving, and I think Oxford Molecular is providing a lead in producing this, is fully networked, enterprise-wide solutions, where any researcher, with very little computer knowledge beyond his Microsoft Office package, can sit down, enter his code word, pull up a menu, go into any level of the system, and perform complex operations of research, retrieval, analysis, hypothesis generation and checking, patent searching, literature searching, report filing and building, editing, and so forth, all on one machine at one time.
BioInform: Does that mean that despite the platform-independent potential systems will start to move toward simpler hardware platforms, like PC's?
Marchington: It all depends on hardware manufacturers. PC's are starting to look more like Unix boxes all the time and vice versa. They're all converging. There is a big performance difference, but if you're on a network you could be working on a PC on the network and you don't need to know what's actually doing the analysis behind that. It could be that much of the analysis is done on the PC locally, but in a certain case, if you ask it to do a certain task, without it's asking you, because it doesn't need your permission, it can dive off into the network and get that performed somewhere else, and come back and say all right, now that's churning away, can I do something else for you in the meantime? And when it's churned away, up pops a window and there it is.
BioInform: What is Oxford Molecular's strategy for finding good people in this tight hiring market?
Marchington: We stay particularly close to our university contacts. So we have our academic contacts in Oxford and Cambridge, people like Phil Green in Seattle, Jim Watson at Cold Spring Harbor, Rich Roberts at New England Biolabs; they to me are one of the most important sources of new talent. We stay close to the bioinformatics schools.