Several years ago, when everyone blithely assumed there were at least 100,000 human genes, Jurgen Drews, the former head of global research for Hoffmann-La Roche, tallied the number of molecular entities targeted by the full armamentarium of drugs that were then on the market. The result was a paltry 482, a far cry from the potential total number of targets, which Drews estimated then at somewhere between 3,000 and 10,000.
Earlier this year, Catherine Burgess and Jim Golden, bioinformatics researchers at CuraGen Corp., revisited Drews' calculations. They concluded that he had actually overestimated the number of existing drug targets. Speaking at the recent IBC Drug Discovery and Technology conference in Boston, Golden reported that, after accounting for ill-defined, redundant, or non-genome-encoded targets, the true number of discrete targets was in fact just 272.
But there was an even bigger surprise in store. Burgess and Golden went on to evaluate the
|If it takes the better part of two decades of intense study to begin to fathom the function of a handful of genes encoded in less than 10,000 bases of HIV DNA, how long will it take to crack the human genome, which is merely 300,000 times larger?
current arsenal of commercially available drugs through the Pharmaprojects database, and proceeded to total the number of drug targets circa 2002. The number had grown — by one. It is stark evidence of the stagnant state of the pharmaceutical industry's pipeline, although, as venture capitalist Steve Burrill has noted, there are several hundred biotechnology-related products working their way through Phase II or Phase III clinical trials, at least some of which, one would assume, have a chance of being approved.
Though the number of discrete drug targets has not grown appreciably over the past few years, the good news (of sorts) is that the total number of human genes has shrunk — most estimates suggest about 30,000 genes — providing the flattering illusion that a greater slice of the human genome is being targeted than was previously believed. Mining the genome is now big business for a host of biopharma companies, including Human Genome Sciences, Incyte Genomics, Celera Genomics, and many others. "The race just started with the completion of the human genome," says Golden, who likes to draw parallels between the genome cartographers of today and the intrepid Portuguese naval explorers of 500 years ago.
The New Cartography
CuraGen's exhaustive survey of the human sequence predicts a total of about 58,000 genes — considerably higher than today's general consensus. "The number of genes is interesting," says Golden, "but as a business, it's how many genes can you do something about."
Burgess' group has identified more than 8,000 potential "druggable" targets, which fall into three categories: 4,990 are potential small-molecule targets, 2,329 are antibody targets, and 794 are targets for protein therapeutics. Discounting gene products that are either patented or in the public domain, they found 2,666 available small-molecule targets, 1,037 potential antibody targets, and 346 protein therapeutic candidates. Of this group, CuraGen has already filed patent applications on around 3,000 pharmaceutically tractable targets, or 75 percent of the available targets it has identified, spending about $500,000 per month in the process, according to Chief Information Officer John Murphy. Of course, the ultimate success of CuraGen's strategy will be judged not in the number of "druggable" candidates submitted to the patent office, but years from now, when we know whether the patents have been approved and survived the inevitable legal challenges.
Doubtless numerous other firms are conducting similar surveys of the druggable genome. In last month's Nature Reviews Drug Discovery, for example, Pfizer Inc. researchers Andrew Hopkins and Colin Groom reported that existing drugs target 399 nonredundant molecular targets that are members of just 130 protein families. Extrapolating from this figure, and analyzing the full panoply of genes in the human genome, Hopkins and Groom predict the existence of precisely 3,051 small-molecule drug targets, significantly less than CuraGen's estimate of closer to 5,000, although the Pfizer figure is based on experimental ligand-binding studies and is thus expected to be lower.
Sequence to Function
As interesting as these calculations are, predicting protein function from gene sequence computationally remains an excruciatingly imprecise science. To the delight of old-guard biochemists everywhere, understanding gene function from raw sequence still requires lashings of good old-fashioned "wet lab" experimentation — and buckets of perseverance.
Take the miniscule genome of HIV — the virus that causes AIDS. Researchers at the University of Pennsylvania School of Medicine recently established the function of a key HIV gene called Vif (virion infectivity factor), revealing a promising new therapeutic target in the process. Good news indeed, but if it takes the better part of two decades of intense study to begin to fathom the function of a handful of genes encoded in less than 10,000 bases of HIV DNA, how long will it take to crack the human genome, which is merely 300,000 times larger?
To be sure, researchers continue to make great strides in unraveling the molecular pathogenesis of genetic diseases. But even when the molecular culprits are finally nailed down, devising an effective therapy will usually take years, perhaps decades. Cures for cystic fibrosis, Huntington's disease, and Alzheimer's disease are still distant dreams, even though the genes responsible have been known for 10 years or more. Then there is still the time required for clinical testing of promising drug candidates.
Despite these hurdles, CuraGen's Golden is in no doubt that deciphering the druggable genome sequence will ultimately restock the therapeutic pipeline. "This is the season for optimism," he insists. "Bioinformatics is the key — it's not just sequence comparison, it is the essence of drug discovery." Time will tell ...
Kevin Davies, Ph.D.