With the complete sequencing of the human genome, one might think that the identification of new potential drug targets would be coming to a halt. Yet there are still many exciting developments to come in the field of target identification, with technological advances enabling challenging biological questions to be addressed in increasingly creative ways.

According to Kevin Fitzgerald, group leader of applied genomics at Bristol-Myers Squibb headquartered in New York, one of the biggest challenges is to understand the underlying physiology of drug targets. “We do not yet understand all of the basic underlying principles and rules,” he says. “We are left trying to interpret and predict the behaviour of dynamic processes based on small snapshots in time. The electrifying part of being involved in biology today is that the technology is beginning to catch up. If you are able to take enough snapshots under different stimuli and conditions, over time, concepts and rules do emerge.”

Kevin Fitzgerald: finding rules. Credit: P. KAYNE

Part of this challenge is to identify interactions between drug targets, which are generally proteins such as receptors and enzymes, and the proteins and other molecules that regulate them. As Jan Mous, president and chief executive of IntegraGen of Evry, France, says: “The most exciting development is also the most challenging: to first understand and appreciate the subtle interaction of many proteins and different messengers and hormones steering the function of our cells and organs, and then pick the best target for drug interaction giving the most favourable ratio of wanted versus unwanted effects.”

Interactions are often subtle and complex, requiring a multidisciplinary approach and the use of informatics systems to make sense of the increasingly complicated data. “A very tight integration of computer sciences and mathematics with experimental biology and chemistry has to be the modus operandi for the future drug-discovery business,” says Mous.

One approach to investigating physiology is through genomics. Now that the human, mouse and rat genomes have been sequenced, high-throughput genomics can come into its own. Ingenium Pharmaceuticals in Martinsried, Germany, uses large-scale chemical mutagenesis and screening to look for novel, medically relevant biological mechanisms in mouse and rat mutants. When a phenotype of interest is found, the mutation is located by high-speed positional cloning and a line of model animals can be produced. Further biological information can be gathered through gene-expression and cell-pathway analysis, complementing the pathophysiology observed in the mutant animal.

A new mechanism of sensitization to the hormones leptin and insulin was recently discovered using this technology. “The effect of the inhibition of the newly discovered pathway is that our mouse model has highly improved tolerance to glucose while requiring only a fraction of insulin for this regulation,” says Ingenium's chief executive Michael Nehls. “In contrast to other insulin-sensitizing effects, the mice do not become obese — on the contrary, they slim under a high-fat diet. We believe this could be a breakthrough in obesity/diabetes research,” says Klaus Dembowsky, vice-president of drug discovery

From genotype to phenotype

What it means to be human? Credit: AFFYMETRIX

Despite the increasing emphasis on proteomics in target identification, DNA microarray technology is still a powerful technique for identifying genes involved in susceptibility to diseases. Target discovery and identification should soon be benefiting from the DNA microarrays of the complete human genome now sold by several companies. First to market in July 2003 was NimbleGen in Madison, Wisconsin, with a chip containing 200,000 probes, with an average of five probes per gene. This was followed in October 2003 by the Human Genome U133 Plus 2.0 array from Affymetrix in Santa Clara, California, with 1.3 million probes and the ability to analyse the expression of about 47,000 different transcripts. The most recent human whole-genome chip, launched last month, comes from Agilent Technologies, in Palo Alto, California. Agilent's double-density-format chip represents about 41,000 genes and, with Agilent's ImaGene image-analysis software, is compatible with most commercial microarray scanners for 25×75-mm chips. This open-platform approach, the company claims, will make it easier for scientists to move from their own homemade chips to the Agilent chip.

Microarrays are already widely used to study gene expression and to detect sequence features such as single-nucleotide polymorphisms (see Nature 422, 917–923; 200310.1038/422917a). Future uses will include assays to study the many forms that a protein target may take as a result of alternative splicing. “For RNA analysis, over the next year, we will start to see assays for efficient splice variant analysis that will enable us to move from gene-specific analysis to transcript-specific analysis,” says David Craford, vice-president of business development at Affymetrix. “More data are emerging to suggest that for many diseases this will be incredibly useful.”

According to Greg Yap, senior marketing director in DNA analysis at Affymetrix, one of the newest and most exciting applications of microarrays to target identification is the possibility of genome-wide association studies of complex genetic diseases. “The experiments that geneticists really want to do are to find genes that cause complex diseases in unrelated clinical populations,” says Yap. “Finding genes that cause a disease in a family makes a nice story, but finding genes causing a disease in the world makes a drug target.” Studying genetically unrelated people is typically difficult, owing to the numbers of genetic markers per person that need to be studied. The advent of microarrays has made individual genotyping feasible, “making it realistic to compare the genomes of people with and without a disease at high-enough resolution to find potential drug targets”, he says.

Pharmaceutical company Roche in Basel, Switzerland, is using the Affymetrix whole-genome expression arrays to identify genes that play key roles in disease pathways. “One approach for new target discovery that we use is to identify the genetic basis for disease susceptibility using a coupled genetic and genomic analysis,” says Gary Peltz, head of genetics and genomics at Roche. “Having the complete genomes of several organisms (human, mouse and rat) available on chips is a major advance. It allows complex biology, especially as it relates to disease processes, to be explored in an efficient manner.” Using a mouse model of a trait associated with a human disease, his group identifies regions of the mouse chromosome that control susceptibility to the disease-associated trait. Gene-expression profiling with the mouse microarrays is also done on target tissues from the mouse strains under study.

“The differentially expressed genes encoded within the identified chromosomal regions are the candidate genetic susceptibility loci,” says Peltz. “We then analyse the function of the candidate genes (and pathways) in the biology related to the disease.” A mouse model of osteoporosis was recently developed using this method, and an enzyme identified that affects bone mass and bone quality, which led to the development of a new potential therapeutic for treating osteoporosis.

Jan Mous: faster mapping.

IntegraGen's GenomeHIP technology platform takes a different microarray-based approach to discovering genes linked to complex disease traits in humans. The technology uses DNA chips for genomic comparisons of pairs of related individuals with the same disease to discover areas in which the genomes are identical — on the assumption that at least one of the relevant disease genes in such individuals will be identical in both. Crucial to IntegraGen's technology is the prior removal of most of the polymorphisms that differentiate the two genomes, cutting down the amount of genome that has to be searched. The company claims that its method is cheaper and faster than conventional methods of gene identification, such as linkage analysis using microsatellite markers. Unlike such methods, which can require several generations of many different families, the GenomeHIP technology can identify potential disease genes with high statistical significance in a relatively small sample of patients (50–150 pairs of related individuals with the same trait) in less than nine months.

Using this platform, IntegraGen recently identified a mutant G-protein-coupled receptor that is strongly associated with obesity in various populations. “The most frequent mutation observed was demonstrated to impair the signalling through this receptor by its natural ligands,” says Mous. “The normal function of this receptor, which is predominantly expressed in the gut, is to transmit an anorexic signal, which could be impaired in patients expressing the mutant form.”

The advent of the complete human-gene chips is generally welcomed as a important tool. But Fitzgerald cautions that, like most molecular tools, gene chips have limitations of which one must be aware. “Experimental design with an eye towards rigorous statistical analysis is crucial,” he says. “Gene chips produce a large number of individual gene patterns and if one is asking a computer program to find patterns in the data, it will find patterns. If the experiments were not correctly designed and controlled, those patterns can be quite misleading.”

He also stresses the importance of remembering that in general, gene chips measure correlative rather than causative events. Andrea Gnirke of Xantos Biomedicine in Munich, Germany, also makes the point that the chips provide an analytical challenge. “Chip data can only be correlative data and implicate an enormous effort for data analysis, and it is difficult to sort out the ‘good’ targets that have causative effects in diseases,” she says.

And because much cellular regulation occurs at the level of proteins, “increased transcript production as measured by gene chips does not always correlate with production of protein, and even if more protein is produced it may not be active because it requires post-translational modification or relocalization within a cell”, cautions Fitzgerald. Jean-Jacques Yarmoff of Hybrigenics agrees: “Modifications of proteins lead to an additional level of complexity that is not necessarily addressable by human-gene chips.”

Tackling the ‘interactome’

Because interactions between proteins are key to their function, determining the pathways within which a potential drug target acts in the cell and the interactions it makes with other proteins are crucial. To learn more about these complex interactions, researchers are starting to map and study the ‘interactome’, the comprehensive listing of which protein interacts with which in an organism. A recent breakthrough by Curagen researchers is the production of an interactome for Drosophila melanogaster, the first comprehensive protein-interaction map for a multicellular organism. They used a high-throughput version of the yeast two-hybrid method to identify more than 20,000 unique interactions involving 7,000 genes in Drosophila. Interactions within and between protein complexes were also studied using mathematical modelling.

The study points to a connection between proteins known to be altered in human diseases and the ‘druggable’ classes of enzymes — those enzymes known to bind and be altered by small-molecule drugs. This is a relief for those trying to find potential therapies for human diseases such as cancer, heart disease and diabetes. Many of the interactions discovered in the Drosophila study will now be examined further in human cells for their relevance to disease. Curagen has obtained a similar set of human protein–protein interaction data by the same methods, but is keeping them under wraps for now.

In another approach to target discovery, Curagen has developed a group of targets it calls the ‘pharmaceutically tractable genome’ (PTG). This is a collection of about 8,000 genes whose products are predicted to make good targets as judged by one of several criteria: they are present in blood or on cell surfaces, rendering them accessible to drugs and antibodies; or are in a class of intracellular proteins that can be modulated by small molecules. This approach allows researchers to focus resources on investigating targets that are both novel and yet likely to lead to successful drugs.

Paris-based Hybrigenics also uses functional genomics for target identification, based on a proprietary yeast two-hybrid method that evaluates which proteins interact with the protein of interest, and whether the interaction is statistically significant. Yarmoff, senior director of business development, says that the company's approach also provides functional information. “Because we obtain many fragments which interact with the yeast two-hybrid bait, we can compute the intersection of these fragments to define the selected interacting domain (SID),” he says. “This is extremely useful information, as by comparing the SID with known functional domain sequences, we can understand not only which protein our bait interacts with, but also the function of the domain at which this interaction takes place. The richness of this information is what allows us to go beyond simple target identification to use this information for target validation.” The large amounts of data generated by this approach can be visualized using Hybrigenics' PIMRider viewer, which enables these protein-interaction networks to be explored.

Channel hopping

Aviva's SealChip16 for patch-clamping (above) is the heart of its PatchXpress7000 system (left) for assaying ion channels.

Ion channels are the basis of neural function and so are attractive drug targets but, historically, screening for drugs that affect ion channels has been a bottleneck in target identification because the tests cannot be automated successfully — the quality of the data obtained has not been good enough. The alternative, electrophysiological screening of individual cells by a researcher, produces high-quality data, but at a pace that is slow by drug-development standards, and is extremely monotonous and tedious to do. Advances in this area are much needed, because although analysis of the human genome has predicted that ion channels could account for up to a quarter of all potential new targets, only 5% of drugs currently on the market affect ion channels.

To address this problem, Aviva Biosciences in San Diego, California, has developed its SealChip16 system. This makes high-throughput, high-quality, patch-clamp recordings from single cells, in conjunction with Aviva's PatchXpress7000, a high-throughput patch-clamp workstation developed in collaboration with Axon Instruments of Foster City, California. Aviva's system incorporates a high-throughput planar patch-clamp technology that gives micro-positioning of cells and tight high-gigaohm seals.

Future challenges

“One of the major challenges in discovery is the complexity of the interactions that result in normal or disease states,” says Seth Cohen of Caliper Life Sciences in Mountain View, California. “In the past, a simplistic view was, by necessity, taken, which resulted in many drugs failing in preclinical or clinical trials for lack of efficacy or side effects. The newer approach must account for this complexity. The holistic systems-biology approach to research will be necessary to overcome this challenge.”

Mous agrees: “A challenging new development in the field of drug-target discovery is systems biology, or the recognition that genes, or better the gene products, are part of, and function, in large complex networks. Understanding the malfunctions in these large pathways will allow the selection of the best drug target for compensating the malfunction of the homeostasis of a biochemical process in a diseased person.”

Peltz stresses that studying target pathways will be critical to identifying a target that is important in disease. “The major challenge is the lack of ‘connectivity’ with pathways for most of the genes in the genome,” he says. Studying the disease pathway is the best way to find targets that are most suitable for drug development, he adds. Unfortunately, this is nearly impossible to do at present, because we do not know enough about most genes' interactions. This lack of information, says Peltz, “makes many of the genes we identify as associated with a disease ‘isolated islands’ that we cannot connect with the ‘mainland’. Hopefully, the protein-interaction network maps that are being prepared in various organisms will help to fill this gap.”