CAMBRIDGE, Mass.--When Cereon was founded early last year in a collaboration between Monsanto and Millennium Pharmaceuticals it gave Monsanto a running start into agricultural genomics discovery by providing the life sciences giant access to technologies, including bioinformatics, that Millennium had already developed for drug discovery. Cereon's mission became to develop core competencies in bioinformatics DNA sequencing, physical mapping, expression profiling, and high-throughput screening.
The deal, which gave Millennium entrée into the ag biotech industry, called for Monsanto to pay Millennium $118 million in upfront licensing and technology transfer fees and additional payments up to $100 million over five years for achieving certain objectives, as well as royalties.
Stanley Letovsky, formerly the principle investigator and informatics director of the Genome Database of human gene mapping data at Johns Hopkins University, is Cereon's director of bioinformatics and coleader of Monsanto's worldwide bioinformatics efforts. BioInform spoke with Letovsky recently about the part bioinformatics plays in agricultural genomics.
BioInform: What is the role of bioinformatics in your agronomics efforts?
Letovsky: The overriding perception within the whole agronomics area is that genomics and bioinformatics are the way ahead. Breeders have managed increases in crop yields of about 2 percent per year for generations, but those gains are becoming increasingly hard to come by. There is general consensus that to maintain this growth more knowledge-intensive methods need to be adopted. In order to improve the traits that farmers worry about, such as increased yields and pest-resistance, we need to understand plants in detail with much more information on metabolic and regulatory processes.
So, there's a rush on right now to understand the genetics of agricultural plants, pathogens, potential symbiants, and model organisms. The other challenge is to figure out how to translate that understanding into product. Without going into too much detail, we are embarked on structural and functional genomics research on a fairly large scale.
BioInform: Are these technologies being used to exploit new traits or to refine traditional ones?
Letovsky: It's a little early to make any predictions. At the moment most of our efforts are channelled into yield improvement and pest resistance, but already products such as herbicide-resistant crops show that the development of entirely new traits is a possibility.
BioInform: How does proteomics figure in your R&D effort?
Letovsky: Proteomics is really a part of functional genomics. RNA expression and protein expression are just two different windows on gene expression. Generally there is a shift going on from structural genomics to functional genomics. The learning curve is in turning functional genomics into a high-throughput operation. The interesting thing is, when you try to find out what genes are doing, you look at RNA and you look at proteins and you get different answers. Scientifically this is interesting but not surprising. The protein concentration must be the time integral of translation, and translation may be fluctuating all over the place. Also, it's not only the rate of production, but the rate of degradation of the protein which is of interest. It's a complicated picture, but once you've got measurements of RNA and protein, this gives you a whole new window of understanding of what's going on in cells and tissues.
BioInform: What are the functions of your bioinformatics staff?
Letovsky: A major function is the mining of high-throughput data generation in a forward processing mode, which essentially is trying to do discoveries. Also, some of our bioinformatics effort is targeted towards maximizing the possibilities of particular discoveries. There's kind of a pipeline, just like in pharma, and as you move down the pipeline things go from a more discovery-oriented high-throughput mode into more focused work on small numbers of genes. Bioinformatics is involved across the whole of that funnelling process.
BioInform: How much do you rely on bioinformatics vendors compared to in-house software development?
Letovsky: We have relationships with many if not most of the bioinfomatics vendors, including Incyte, Pangea, Molecular Applications Group, IBM, Silicon Graphics, and Proteome Systems, while discussions are going on with several other vendors. We hedge our bets, doing a certain amount of in-house development and integration, keeping our eye on software house developments and adopting their solutions opportunistically. One of the problems is that there are a lot of software houses, probably too many. Not all of these bioinformatics companies who are hoping to sell their products to the same 200 customers will generate enough income to survive--there's likely to be a big shakeout over the next couple of years.
BioInform: What are the biggest bioinformatics challenges for Cereon?
Letovsky: A major challenge is to integrate these relationships effectively. One problem is that every vendor wants to sell us an end-to-end solution whereas, in principle, we would much prefer to be able to mix and match solutions, taking components from each of them and integrating them. There is also an issue about whether the industry really has the incentive to develop a component model; it's a problem when you've got a big fat database in the middle of the system. Clearly some people have been trying to push this, but it's a difficult question to get right. We can build up software from components, but we can't say here's a data module that you can plug into your database.
Another problem area is developing research software to tight timelines. Software engineers are used to working on product cycles that range from nine months to two years, but we really have to think in terms of delivering functionality on a much shorter time-scale if we don't want software to be rate limiting.
BioInform: What are the technological challenges of genomics-based product discovery that cannot be solved by off-the-shelf bioinformatics tools?
Letovsky: First of all, I would dispute the notion that there are very many off-the-shelf bioinformatics tools. A lot of these companies are selling software that is still under development. Relationships are cooperative, halfway between codevelopment and buying a product.
An area of sequence analysis that is fairly mature is the zone of greater than 50 percent homology. Then there's another area where the homologies are more distant, a twilight zone where people bring a lot of extra computing power to bear. There is still a lot of method development in that area.
There's a whole other area of inferring function from homology and one of the big problems right now is that there is no standard within the community for describing functions of proteins. There's not much control over how functions are expressed, it's pretty haphazard. There's a lot of work going into developing better standards for describing molecular functions and biological functions and a number of different categories will need to be addressed.
In functional genomics there is also a big effort going into finding out what are the best algorithms for pulling useful information out of expression profile data. It's wrong to talk about off-the-shelf solutions here as it is not a mature problem at all. The algorithms are going to vary as the sensitivity of the wet lab methods vary. One of the areas of concern to us is the scarcity of public or private-sector databases of pathways and regulatory mechanisms. There are a few of them but because they are so labor-intensive to compile, not nearly enough.
BioInform: How do you see things shaping up in general with respect to bioinformatics, in the medium to long term?
Letovsky: I think that agricultural bioinformatics is at a very exciting stage right now. There are lots of job opportunities and lots of interesting research to be done. In agriculture we tend to be interested in a wider range of species than in pharmaceuticals, so the focus on comparative biology is stronger. The research being done over the next 5 to 10 years has a good chance of enabling another green revolution that will allow agriculture to keep pace with population growth for the next half century, with reduced environmental impact. One of our biggest challenges at present is to convince consumers and voters that genetic improvement of crop plants is our best hope for more sustainable, higher yielding, and less chemical-intensive agriculture.