Function genomics is the study of individual genes, proteins, or pathways within the broad context of the genomics of the cell, tissue or organisms. Specific functions are considered within a large percentage or even the entirety of the genome of that cell.

Functional genomics is important for the new biological insights it provides and because it may yield the next major discoveries in pharmaceutical therapeutics. Genomics will shorten the painstaking 10-year process usually required for lead optimization to about 3-4 years. Using the genome will shorten the process of target identification (characterizing genes as candidates, using disease tissue, cellular models, and animal models), target validation (discrimination of valid targets relevant to the disease), lead identification (using compound screening with high throughput screening) and lead optimization (search for optimal efficacy in further studies or clinical studies).

The Human Genome Project encompasses a large number of goals, particularly the genetic and physical map that is nearing completion. A great effort has also been made in the areas of technology transfer and informatics, which will be the keys over the next decade to exploiting all of this information. Efforts have also been made in the areas of education and ethics. The Human Genome Project began in 1990 and is now an international collaborative year. The project goals are to:

• Identify all (> 100,000 genes) in human DNA
• Determine the 3 billion chemical base pairs of human DNA
• Store this information in public databases
• Develop tools for data management and analysis
• Address the ethical, legal and social issues

The Human Genome Project reported in Nature in December 1999 the sequencing of the 33.5 million base pairs of chromosome 22. They reported there were more than 545 genes, more than 134 pseudogenes, 39% of the DNA was copied into RNA, only 3% was made into protein, 247 known genes, 150 genes had some homology to genes in the database, and 148 expressed sequence tags. As of April 2000, 65% of the genome is sequenced and a 90% rough draft will be available in June 2000, with completion by the end of 2000.

A large number of expressed sequence tags (EST) or cDNA tags are now stored in databases. In the public database there are about 70,000 EST and the proprietary Incyte database claims to have about 90% of the genome completed. It is likely that about 10% of EST are missed in the gene libraries because some genes are probably expressed in very low abundance and some are temporally expressed only in specific tissue. This is likely to be an important factor as it is quite possible that therapeutic targets will be rare genes or temporally expressed genes, because therapies directed at such genes may have the least side effects. Therefore, while there is a bias towards knowing more in the early period, there is also a bias towards possible golden material coming at the end of the EST database.

The number of gene patents in the US alone is skyrocketing, due to efforts to protect various amounts of data. The uses of EST have ranged from efforts to patent the individual sequence alone, to more recent efforts to obtain a patent for the function and possible use of the EST information. Interestingly, while the patents are increasing, the function of only about 10-15% of the current EST database is known. Rough homologies can be drawn for about 50% of the genes, but at least 50% of the genes have no known function.

The dense genetic map and the advent of single nucleotide polymorphisms available throughout the genome will speed the identification of the gene of interest. This will simplify the genetic analysis of complex, polygenic diseases, which was previously accomplished by positional cloning.

Although epidemiologists and molecular epidemiologists are excited about this new information, it must be remembered that even with this new ability most of the genes will encode for proteins for which there is no known function, as illustrated in Figure 1. If this is true, it will be important that much of the expression pattern data is available to the public to narrow the analysis to the specific gene that may be causative in a disease by looking at the information related to the gene expressed in the disease or tissue of interest. Expression patterning will work together with genetic mapping to allow faster identification of gene function.

Tissue expression methodology

Past methods for defining expression profiles have been slow and tedious, including the use of subtractive hybridization, differential display, and mass sequencing and serial analysis of gene expression techniques. This will be replaced by hybridization arrays, which are the most efficient and will be the least expensive ways to perform expression profiling in the laboratory.

A genetic microarray is the analysis of up to 20,000 genes simultaneously. In the Synteni technique one form of RNA is labeled with a red fluorescent marker and the other with green which are then hybridized and read by a laser scan for the differential expression. There are internal controls, but there is a single hybridization for each single gene of interest. A DNA microarray reproducibility experiment shows that the microarray is equivalent to performing a Northern Blot on 10,000 genes simultaneously (Fig. 2). The reproducibility has been quite good and the quantitation has been remarkable, particularly compared with techniques such as differential display.

The Affymetrix system is a fascinating system in which photolithography is used to synthesize the oligonucleotide sequences on the chip in solid phase. It is very sensitive and specific quantitatively, which is important as about 20 hybridizations are done for each gene of interest and 20 controls. The advantages of this system is that it is truly "plug and play", extremely specific, highly sensitive, and interfaces with bioinformatics. It is high density, due to the photolithography technique, and can analyze about 100,000 genes per single chip. By the time the genome sequencing is completed, there is the potential for analyzing every gene on one chip. The disadvantages are that making the photolithography mask is expensive and tedious, making the chip itself very expensive. The cost per chip has fallen from about $10,000 to $1000-2000. The Affymetrix chip can not be customized for a single laboratory.

In-house development of chips is now possible due to an evolution and explosion of the technology. Individual laboratories may find this daunting, but individual research centers can undertake this. About 20 different companies provide the technologies to make this possible. The advantages are lower production costs (after the cDNA purchase) and custom design to a particular diagnostic problem. Disadvantages are the initial set-up, the initial cost of the cDNA, errors in EST database or cDNA suppliers that may require mass sequencing and validation steps, and a loss of sensitivity and specificity compared to the Affymetrix technique.

Tissue expression will have remarkable utility. Tissues or cells may be input to look at the effects of compounds or inhibitors or specific mutations that may participate in pathways, resulting in outputs that may include prognosis for cancer, diagnosis of different types of tumors, therapeutic targets for drug validation, efficacy, and toxicity.

Pharmaceutical companies are beginning to use this ability to determine toxicity by mapping out the hepatic pathways for different types of toxins. With this technology it is possible to visualize that particular compounds are toxic in rat livers at a very early stage, for example, well before extensive toxicology studies with hundreds of animals. In these days of mass high throughput chemical screening this analysis is very useful before spending lots of money.

Application in cardiovascular medicine

Lee used transcriptional profiling, or transcript imaging, to map a picture of an area that is not well understood: the molecular response of cells in human atherosclerotic lesion. The results obtained were highly specific, all of which were confirmed by multiple Western and Northern analyses. Clusters of functions seemed to be induced by the particular stimuli. The small number of genes induced simplified potential identification of physiologically relevant genes. This was a special circumstance, as usually in the lab a larger portion of the genome is being affected.

In this work, Lee looked at different pathways that might affect atherosclerotic lesion stability in human aortic smooth muscle cells using genomic screening. Free stimuli of interferon, tumor necrosis factor (TNF) and mechanical strain were studied using Northern blot analysis as a control (Fig. 3). MHC II was robustly induced by interferon, TNF induced MMP-1, and vascular endothelial growth factor (VEGF) was mechanically induced. Good internal controls are needed with these genetic studies to save time and money and confirm good specimens. In the 10,000 gene array, dozens of genes were induced by TNF, including a number that were already known such as MCP-1, superoxide dysmutase, IL-4, VCAM, and ICAM-1.

Interferon gamma was even more prolific and induced a large number of genes, many known to be involved in vascular biology in response to interferon and a number involved with interferon-induced apoptosis in other cells (MHC Class II, ICAM-1, MCP-1, SOD, and ICE). Interestingly, mechanical deformation led to a very small number of induced genes, compared to the interferon and growth factor stimuli (Fig. 4). The types of genes induced were quite specific. There was downregulation of MMP-1 (leading to collagen accumulation) and upregulation of PAI-1 (leading to decreased matrix degradation); both genes participate in intracellular matrix degradation. Also present was upregulation of VEGF and upregulation of cyclooxygenase-1 leading to increased prostacyclin.

Other use of transcriptional profiling include quality controls for primary cell populations to rule out false positives and determine reproducibility. One of the advances of bioinformatics is performing mathematical clustering techniques using the same types of analyses used for multivariate epidemiology. Those techniques are moving closer to being able to identify common pathways between cells. In this fashion a molecular map can be developed in which all information is integrated through the genome into common pathways that can be understood. This illustrates the need for information to be shared and why information technology is needed to share information from the different types of experiments.

Expression patterns will be pursued, and meanwhile classical biochemistry and cell biology will continue as it is the basis. In vivo validation using transgenics and knock-outs is a critical factor. Exploiting gene transfer and integrative physiology further with techniques such as high throughput yeast to hybrid screenings is needed.

A summary of the known protein or gene targets for all of the therapies currently in use shows there is about 483 targets. About 45% of the targets are receptors and 28% are enzymes. The pharmaceutical industry estimates using high throughput screening there are 4000-4500 different small molecule genes that are targetable. Over the next few years the data to increase these targets 10-fold will become available. This ability is unprecedented and an exciting part of the entire genome project.

Microarray research must be just as robust and thoughtful as classical hypothesis-driven research. It can not rescue a bad model, bad experiment or design. With microarray research an hypothesis is researched across a large portion of or the entire genome. It is important to consider that taking great care in these experiments can save years of chasing false leads, Good controls, a great deal of care, reproducibility, the same sort of thought that goes into all experiments, are the key items in microarray experiments as well.