Introduction

The discovery of new chemical molecules or novel medicines for the treatment or cure of human diseases requires a broad range of scientific and clinical expertise. As the whole human genome is about to be deciphered and is composed of three billion DNA base pairs encoding about 120-140,000 proteins, the post-genomic research area challenge is to understand the functions of the protein targets that are causing or interfering with a disease state.

In the last decade, numerous technologies have been developed by the pharmaceutical and biotech industries to identify proteins (about 5-10,000) potentially involved in major human diseases.

Target validation represents a research activity that allows a reduction of the number of potential targets from 120,000 to a more realistic number (a couple of dozen) for studying a particular disease, and to initiate a research program to discover new drugs or medicines. Filtering or funneling research activities--such as human genetic linkage analyses, animal models, or in vivo cell-based validation--are used extensively to choose the right targets. Since the drug discovery process is very long and complex (from a gene/protein to a chemical molecule on the market it takes about 10-12 years), only a very small number of targets can be develop at any given time, even in large pharmaceutical companies. Therefore there is a great need to discover and focus on technologies/discoveries that can speed the early stages of target identification/validation and that allow the association or linkage of a given class of proteins, genes, or pathways with a specific human disease. It also helps save critical time and resources by selecting appropriate targets for further development at very early stages.

Novel approaches to deal with numerous potential targets at once (high throughput) or in parallel in order to select for the best ones are desperately needed for efficient drug discovery. Bioinformatics, in vitro and in vivo assay development, ultrahigh-throughput devices, microarrays, antisense, protein-protein interactions, and proteomic technologies are among the key elements for a successful approach in early drug development.

Recent Development

A recently held meeting in San Diego, organized by the IBC USA Conferences, provided a pertinent survey of currently available technologies for bridging the gaps between the explosion of sequence information and unknown protein functions or linkages with disease states. The program covered current trends of target validation technologies and pertinent issues such as what model organisms to use for targets/pathways validation (P. Nef, Hoffmann-La Roche, Ltd.), the use of antisense to deal with multitude of genes at once (K. Giese; Atugen; T. Woolf, Sequitur), proteomic approaches to deal with the poor correlation of expression between mRNA and protein levels (C. Rohlff, Oxford GlycoSciences), or differential display (D. Lewin, Curagen) between normal and disease sates using gene profiling with microarrays (J. Reteifs, Affymetrix), with cDNA expression libraries and databases (L. Arnold, Incyte), or fluorescent biosensor reporters randomly integrated into the genome (G. Henkel, Aurora). Specific examples were also presented, such as the validation of anti-infective agents in vivo using intracellular peptide presentation (S. Shimer, Cubist), or a way in which disease-inducing genes can be identified using laser capture microdissection of control versus transformed cells (F. Randazzo, Chiron), and a novel reverse genomics approach to validate genes involved in malignant cell transformation (J. Barber, Immusol).

The main conference was preceded by a workshop on the use of bioinformatic tools to navigate the genomic data explosion (D. Davison, BMS Research Institute) and new ways to analyze genes and proteins sequence relationships (S. Benner, EraGen). Experimental approaches using genetic material (QTL analysis in murine or human subjects) to directly identify novel targets or altered pathways causing a disease were not presented. Traditional approaches to identify the function of a cellular target after exploring the binding patterns of natural products or other ligands to the novel target were presented only en passant.

Target Validation in the Future

Bioinformatics is recognized as an essential tool for dealing with complete genome sequences, for extracting correct gene coding sequences and the corresponding proteins, and for allowing multisequence comparisons across several species at once. The complexity and size of such nucleic acid and protein sequences, plus the growing number of EST and SNP data is enormous. New powerful hardware and software are needed to link databases via an efficient network and to mine these relationship databases with automated algorithms, 24 hours a day, seven days a week to provide up-to-date information to biologists, drug designers, and clinical MDs. Virtual screening of ligand against newly annotated protein targets is likely to help drug design using three-dimensional models of protein motifs.

Proteomics is helping to identify targets or marker proteins causing or specific to a human disease pathogenesis via serial measurements from accessible compartments (e.g., body fluids such as serum, urine, synovial fluid, CSF). Although rather technology limited by the number and types of protein classes that can be analyzed at once, it is a valuable technology for soluble, cytoplasmic proteins, whereas membrane-bound proteins are much more difficult to handle and analyze with this technology. Normal proteomic analysis requires sample protein separation on high-resolution SDS-PAGE (one or two dimension) gels or HPLC, spot detection by protein dyes or fluorescence, image and database driven spot selections, to a fully automated robotic processing to obtain peptide sequences from gels with MALDI-TOF and tandem mass spectrometry. The peptide sequences are then stored into databases and compared with existing EST and DNA databases. It is becoming clear that the specific removal of abundant proteins such HSA, IgG, Hptg, or Trf by immunoaffinity unmask underlying proteins from CSF samples, therefore improving the likelihood of finding appropriate markers from "enriched samples" containing up to 2,000 individual proteins (Oxford GlycoSciences). Biomarkers associated with a specific disease state or a sub-population of patients could help define complex traits contributing to the overall clinical pathology. Functional proteomics of protein kinase and phosphatases to examine protein phosphorylation patterns (phosphoprotein maps), signaling pathways and clinically relevant targets have already been applied to the fields of diabetes, liver cancer, and colon tumors (Kinetek).

Microarrays in the form of high-density oligonucleotides or cDNA are expensive but very valuable research tools as they permit one to get a comprehensive snapshot of gene expression in an organism in one experiment. Gene expression profiling and by inference biochemical pathways, are used to understand drug treatments, or disease progression. Without an appropriate software package to handle the millions of data points obtained by image scanning of the chip surface, it is virtually impossible to extract useful information. As an example, the high-density array based on the completely sequenced yeast genome contains approximately 7,000 genes and allows the screening of all ORFs in a single experiment (Affymetrix). It usually requires months of computer analysis to conclude from an experiment. The ideal drug effect should be strong on a specific pathway involved in the disease, and have minimal effect on other pathways not relevant to the pathology. Toxicogenomics is an emerging approach using microarrays to determine at early stage of drug development the potential side effects and safety issues associated with a clinical candidate drug. Novel custom-tailored cDNA microarrays can be purchased and designed according to the customer need. The sensitivity of this technology is increasing to a 2 pg limit and 1.4-fold change detection, with a sensitivity of one in 100,000. As the number of arrays is increasing, databases are available to screen for gene level expression across different set of experimental conditions (Incyte).

Antisense technologies allow to bloc mRNA translation or mRNA stability in a very specific, and high-throughput (up to 500 targets) manner both in vitro and in vivo. Through diverse delivery vehicles (e.g., lipid, virus), short modified oligonucleotides can be inserted into a mammalian cell to inhibit protein expression by as much as 85–95%. Such an approach has been used successfully to demonstrate the role of beta-secretase as a valid target for Alzheimer (Sequitur). If the gene target is down-regulated in a disease state, it cannot be enhanced by antisense. In principle, every up-regulated gene known to be involved in a pathological state can be brought to normal level using antisense. Examples of timely preformed inhibition experiments on the inhibition of blood vessel growth in the eye (rat corneal angiogenesis model), defective mouse development (inhibition of VEGF embryonic expression), and the prevention of tumor growth (Atugen) have already been obtained.

Animal models have been extensively used to understand and mimic human pathologies. Bacteria, fungi, nematodes, fish, amphibians, insects, birds, rodents, rabbits, carnivores, nonprimates, or primates are often used during the different stages of the drug discovery process. Dependent on specific target validation criteria, one model organism will be favored versus another. Antibiotic and anti-infective drug targets are likely to be validated in lower organisms, and a drug safety profile may be determined in rodent or primates to ensure that no side effects are associated with a specific chemical class. However, to mimic complex human diseases such as Alzheimer's disease, Parkinson's disease, schizophrenia, or osteoporosis, the model organism of choice is still the mouse as it can be genetically modified by knock-in or knock-out gene transfer. An interesting possibility is the use of the nematode C. elegans as a high-throughput model system to select and identify both drug targets and pathways, but also chemical structures from large (500,000-4,000,000) compound libraries.

Summary

With the human genome composed of approximately 140,000 genes, and many more protein variants, gene profiling experiments using microarrays are providing a set of about 10,000 potential candidate genes. Adding differential display approaches could reduce the number of targets to 500, or to 50 if combined with a powerful bioinformatic analysis. Applying high-throughput antisense technologies can reduce it further to four or five targets in a timely and economical manner. However, if enhancers or down-regulated genes are the final targets, transgenic studies in mouse or recombinant expression in cell lines for phenotypic studies should be performed, although it represents a very tedious, long, and slow process (very low throughput). It is therefore urgent to invent high throughput knock-out technologies in rodents in order to address the functional roles of many proteins in model organisms related to human pathologies. To rationally reduce the plethora of putative targets represents an intellectual and technological challenge for both pharmaceutical and biotech industries. It is likely that appropriate partnering will be an essential step for the rapid identification and validation of novel promising drug targets to cure major human diseases.

This article was written by Patrick Nef, PhD, Head of Molecular Neurobiology, F. Hoffmann-La Roche, Building 68, room 450a, CH-4070 Basel, Switzerland. He can be contacted by phone at +41 61 68 80 828, by fax at +41 61 68 81 720, or by e-mail at Patrick.Nef@roche.com.

For more information about Drug & Market Development Publications, contact us at +1-508-616-5566, +1-508-616-5544 (fax), http://www.drugandmarket.com or lwood@drugandmarket.com.

To make any comments on this article, or to ask a question of the author, please contact the publisher. If you would like to submit an article, please contact the editors.

The opinions expressed in the articles published in this section do not necessarily reflect those of Pharmalicensing or UTEK Corporation. No actions including proposals to or agreements with other companies should be taken by any reader without obtaining specific business or legal advice. Neither the publisher nor the authors accept any liability for any actions or activities undertaken by any reader or other third party as a consequence of these articles or for any errors or omissions therein.