New technologies and new approaches hone in on better drug target identification, but financial support for discovery remains a challenge.
What does the brain of a schizophrenic person have in common with the muscle of a diabetic person? The answer to that question is an exciting new breakthrough in the search for new drug targets for schizophrenia, brought about by means of transcriptional profiling. Scientists from Psychiatric Genomics, Inc. (formerly of Gaithersburg, Md.) observed that hippocampal neurons from the brains of schizophrenic post-mortem cases had many of the same gene changes as other investigators had reported in diabetic rat muscle tissue, indicating a possible state of metabolic hypofunction and a new set of drug targets for the disease. (Altar et al. Biol Psychiatry. 2008 Oct 29. [Epub ahead of print].) These new drug targets are desperately needed in a time when psychiatric and neurodegenerative diseases are the second leading cause of morbidity and premature mortality in the US.
Missing the target
Pharmaceutical companies have not had the best luck finding new drug targets. Historically, many drug candidates have been identified by serendipity, not by any rational or systematic search. Serendipity can only go so far, and so in the 1990s, drug companies invested millions in high throughput screening programs, intended to test as many drug compounds as possible against as many drug targets as possible. That strategy did not yield the hoped-for results. One of many problems with the HTS approach was the inadequacy of drug targets, particularly the one disease/one target assumption. Now, pharmaceutical discovery is going “back to the drawing board”, hoping to use rational, systematic methods to increase the odds of having a bit of old-fashioned good luck.
The transcriptional profiling approach taken by Tony Altar, PhD, owner of Neurodrug Consulting and former CSO of Psychiatric Genomics looks at RNA abundance, not DNA, and that may make all the difference in terms of success. Some scientists are concerned that the risk assessment values for schizophrenia that are based on SNPs, even for the most significant genes reported to date, do little to predict schizophrenia in an individual or even in a large group of patients. Regarding SNP studies in CNS research, Tony Altar comments: “The very small heritability for psychiatric diseases contributed by any one SNP underscores the fact that multiple genes probably collaborate to produce them.”
Crunching the data
The sciences of genomics and proteomics have given rise to a great many strategies and technologies for finding new drug targets in the laboratory, but one increasingly important tool does not require wearing latex gloves—the computer. With the sheer amount of data generated by the simplest of genomic profiling experiments, for example, powerful computational solutions have now become a necessary part of discovery research. Those tools range from desktop data analysis to sophisticated systems modeling using supercomputers.
Tibco Spotfire (Somerville, Mass.) offers data analysis software that is described as data-agnostic. Spotfire products have broad applications across many industries, because rather than being designed for a specific application, it is based on knowledge of human-computer interaction. Spotfire presents the data in a graphical user interface that can be manipulated intuitively. This type of software can help investigators visualize complex gene interaction networks, or sort out statistically-relevant information from background noise.
Says Mark Demesmaeker, PhD, “People doing the lab work are not necessarily the ones that have the capacity or desire to get sensible information out of the labs they run. … they get a large volume of data out of experiments—microarrays or target identification technology—but once they generate the data, they realize the amount is huge. How do you put it in context?”
The very model of a modern biological
Model building is another bioinformatics approach that can be used in a powerful way for drug target identification. Lukasz Kurgan, PhD, associate professor at the University of Alberta, Canada, is part of an interdisciplinary team that uses digital signal processing (DSP) of the protein sequence to predict ligand binding “hot spots” in tubulin. Because it is involved in cell division, tubulin is a target of interest for cancer, and any of these binding “hot spots” could potentially be exploited as a drug target.
Kurgan describes the DSP model as “a little bit of biology with a little bit of engineering in it.” It uses known properties of amino acids, such as the tendency for binding sites to be flanked with proline, as well as the energies of individual electrons in the molecule, which contribute to the protein’s characteristic frequency. Says Kurgan, “That kind of methodology is built upon the premise that you can find regularities in the data, and out of those regularities you can build a model that will find similar regularities in unknown data.”
In silico modeling is another area of bioinformatics that can be used to improve the efficiency of target discovery. Gene Network Sciences (Cambridge, Mass.) provides computer models from a systems biology perspective. The models are based on Bayesian Interference, which is a probabalistic approach to modeling rather than a mechanistic approach. It produces a large number of possible models.
The advantage of having a plurality of models, sometimes with conflicting predictions, is obviously not in having one clear answer to a biological problem. Systems biology thrives on the generation of multiple hypotheses, iterative modeling, and experimentation. In the case of drug target identification, a more robust process at the hypothesis stage could save millions of dollars in development later. According to Thomas A. Neyarapally, vice president of corporate strategy and intellectual property for Gene Network Sciences, continued dismal clinical trials success rates are evidence that the early discovery process is broken and has not been alleviated by the marked increase in ‘omics and related data and its cataloging in sophisticated databases. In silico modeling can potentially illuminate the biology in such a way that doomed programs can be terminated sooner. Says Neyarapally: “We work with partners to carefully construct experiments designed to shine a flashlight on previously undiscovered biology relevant to the drug or disease of interest. The new biology is captured in the resulting reverse-engineered models, which we probe for novel biomarkers and targets.”
Most drug companies now have reams of data and reports on side effects and adverse events that they have accumulated in the course of testing hundreds or thousands of drugs in the clinic. A new concept in drug target discovery leverages this wealth of safety data to track down new indications for drugs (either failed or approved), and to find new disease targets. The classic example is the accidental discovery of Viagra during the screening of a routine blood pressure drug target. Viagra’s primary indication was a puzzling side effect of a blood pressure candidate in Pfizer’s pipeline.
Publicly-available data from other companies can be used the same way, and many of the bioinformatics approaches being used to search for de novo drug targets can also be used for this “backward searching” based on side effect profiling. This approach is described in a recent article. (Campillos et al. Science 11 July 2008: Vol. 321. no. 5886, pp. 263-266.)
The success of new drugs hinges on the quality of the target or targets. A good target will be strongly-associated with a disease state, well-characterized biologically, and have a powerful associated biomarker. New drug targets can come genomic or proteomic studies, metabolomics, or even from inside a computer. Unfortunately, translational research is suffering from a terminal lack of funding. Without an influx of the most important discovery technology of all—money—pharmaceutical offerings will continue to be pale imitations of past successes.
About the Author
Catherine Shaffer is a freelance science writer specializing in biotechnology and related disciplines with a background in laboratory research in the pharmaceutical industry.
This article was published in Drug Discovery & Development magazine: Vol. 11, No. 12, December, 2008, pp. 29-31.
Filed Under: Genomics/Proteomics