Whither Bioinformatics?
by Richard P. Grant, M.A., Ph.D.

When I started my doctorate in 1991, I was granted access to the University's VAX, with some elementary sequence analysis programs. In the following months our department began running bioinformatic/sequence analysis programs on VAX/VMS for the University and attached groups. The programs available were very basic -- we could assemble sequences, run FASTA and do pair-wise comparisons and hydropathy plots, and very little else. There was great excitement when BLAST arrived, with its real-time searching; which meant we'd spend five or ten minutes sitting looking at some dots on the screen rather than setting up an overnight FASTA and coming back to it when it was complete. There was similar excitement -- not to mention a fair bit of opposition -- when we switched to Digital Unix.

More than ten years down the line, things have not changed all that much. Most people still perform BLAST searches with the default settings, sequences are usually submitted to a remote server, and hydropathy plots are the first step in protein sequence analysis. Of course, there are better algorithms and methods for secondary structure analysis and threading methods for tertiary structure predictions. The total size of searchable sequences has increased tremendously -- we are living in the post (human) genome era after all -- but so has processing power. Thus although the extent of a database query has exploded, the actual time taken has probably decreased. Additionally, the power of desktop computers is such that it is becoming feasible to perform sequence searches on them (see, for example, http://www.apple.com/pr/library/2002/feb/07blast.html and
http://www.edinburgh-biocomputing.com/EBS/download.html). Distributed projects and open source collaborations abound, but the major problem is that the standard sequence comparison and data mining analyses will all be completed in the near future. All genes will have been compared against all known genomes, and -- in a finite time -- all polypeptides will have had their representative structures predicted, if not solved. As the human genome project moved from innovation to brute force number (or base) crunching, from research to support, so the discipline of bioinformatics is changing. As the human genome project has (nearly) run its course and researchers are now looking for the next step, so for bioinformatics.

Biology itself has changed from a dogmatic, 'disciplinary' or 'pathway-based' science, to a broader, multi-disciplinary exercise. According to Shankar Subramaniam of UC San Diego, there is a new 'central dogma'; genomes code for gene products, whose structures and functions are embedded in the pathways and physiology of biological activity. Each metabolic pathway can no longer be considered in isolation, but in the context of the interlocking and cross-coupled networks in which each component of that pathway participates. Thus scientists are leaving simple reductionism (which has until now served us extremely well) behind and arriving at the idea of 'high through-put biology', in which complex biological systems are studied in terms of all their elements -- all genes, all proteins -- simultaneously.

According to Leroy Hood, who created the Institute for Systems Biology in Seattle, such an approach not only needs a greater infrastructure (DNA/gene expression array technologies, proteomics, multiparameter cell sorting, mass spectroscopy, single cell assays, etc) than traditional disciplines (molecular/cellular biology, biochemistry), but also requires advanced computational technologies. The challenge to systems biology includes interpreting genomes, mapping genotype-phenotype relationships, dissecting whole organ function, and determination of complex traits and polymorphism. Aside from expanding currently available methods for whole and partial genome comparison, the 'new biology' will require algorithms to allow metabolic pathway reconstruction, more powerful metrics for similarity and homology, the ability to dissect physiological pathways and a framework on which models may be built. Eventually, the aim is to model all the processes that occur within a cell, tissue, organ or (ultimately) a whole organism.

The contribution of models arising from systems biology (i.e., simulated biology) to human health can not be overestimated. Pharmacogenomics -- the use of an individual's genetic information to determine a personalized drug regime with minimal side-effects for greatest efficacy -- will be revolutionized. More far-reaching will be the ability to investigate and treat any disease using models on a computer -- the 'in silico man'. For any given pathological condition or disease state this ultimately will reduce the time taken for initial screening of candidate pharmaceutical compounds, obviate the need for animal testing, and eliminate clinical trials. Apart from the obvious effect on human health, pharmaceutical companies will be able to bring new drugs to market in a drastically reduced time, squeeze more life out of each new patent and make ever larger profits. There are also endless possibilities for education and entertainment. Imagine what the Natural History Museum could do with these models!

There are two major obstacles to realizing this dream. The first is money, and the second is personnel. Given the above it is perhaps surprising that pharmaceutical companies are not investing heavily in systems biology. It takes visionaries -- such as Hood and Subramaniam -- to get the ball rolling. The systems-based approach does not fit easily into the classical, department-based, academic infrastructure for doing science. A true multi-disciplinary attitude is required, and high-throughput biology needs to marry with sophisticated computational facilities. None of this attracts the traditional academic funding sources. Even venture capitalists, who are not averse to high risk investment, shy away from the length of the funding cycle today; the ten-plus years to get a drug to market is probably going to look positively ephemeral compared with the length of time it will take to have a working in silico man. And as interesting as the steps on the way - the in silico yeast or in silico nematode -- may be, they are unlikely to be great money-spinners.

The US National Science Foundation has recently offered US$8 million in grants for bioinformatics research, through its cross-disciplinary Biological Databases and Informatics programme, which is designed to encourage new approaches to knowledge management. This is a step in the right direction, but is little more than that.

Second, as Shane Sturrock (CEO at Edinburgh Biocomputing Systems) said to me a week or two back, "Where are the bioinformaticians?" There is no lack of skilled programmers, and a fair number of extremely good biologists. Skilled programmers will be in great demand to come up with the sophisticated algorithms and applications that will be required. Armies of biologists will be trained in the new data collection and analysis techniques. Data control systems will be designed and implemented. But that's not the problem. It is more difficult for one with a computational background to understand and know ('grok') biology than it is for biologists to learn to code, but for whatever reason there are relatively few biologists who have cared to take this step. And the instinct for biology needs to be married with the ability to understand and attack problems from a computational standpoint. I am not suggesting that every biological scientist should learn C++ or even PERL, but the success of systems biology will depend on the polymath. Of course, given the right environment, intimate working relationships between computational scientists and biologists may achieve the same result, but this will depend on the setting up of many more Institutes for Systems Biology.

This is not the next big thing in bioinformatics (http://genehack.com/archives/feb2002.html#1014264956). This is what bioinformatics is. It is the goal, the destination, the raison d'etre. To achieve it, we are going to need a huge investment in scientific infrastructure, and the attitudes of many in the profession will need to be challenged. It is time for those with the relevant connections to consider seriously what they can do, and for trained biologists to think carefully about the future, and their role in it.

###

Richard P. Grant, M.A., Ph.D.
Medical Research Council Laboratory of Molecular Biology
Steering Committee Member, The Science Advisory Board

###

[ View Next Perspective ]

[ View Previous Perspective ]