The hard cell: Genomics and the future of drug discovery

From the point of view of a biomedical scientist working to translate genetic discoveries into biological and therapeutic advances, scientific history can be divided into before genome and after genome. Before the completion of the Human Genome Project (HGP) in April 2003, we had a keyhole view of biology. With the completion of the HGP and the advent of its follow-on translational initiatives, for the first time we have a ‘parts list’ of human development and physiology being laid out before us.

It is important to remember that the completion of the HGP marked the beginning, not the end of the ‘genome era’. The term ‘post-genomic’, frequently used to describe the current era of research, implies that we understand all there is to know about the human genome and have applied it to understand biology and disease. On the contrary, although the HGP delivered the sequence of a reference genome, it delivered neither the sequence differences among individuals that contribute to health and disease, nor the understanding of how genes function to produce cellular and organismal physiology. It is this misunderstanding that led to the oversized expectations of the HGP for drug discovery. The genome is indeed advancing biomedical science at a startling rate, and has enormous potential to transform drug discovery. But this potential will require translational efforts every bit as ambitious as the HGP itself. The NHGRI, which led NIH’s contribution to the Human Genome Project, is engaged in just these translational initiatives.

One of these projects, the International Haplotype Map Project, or ‘HapMap’ (see box), is characterizing all the positions in the human genome where our sequences vary – 99.9 percent of everybody’s genomes are the same but 0.1 percent is different, which, considering the 3 billion letters in the human genome, works out to be over three million differences in the DNA sequence. It is these differences that will contribute to susceptibility to common diseases such as cancer and heart disease, and will help determine which people respond to which pharmaceuticals. Until now, we have had no complete atlas of common sequence variants; with the HapMap, we do. Another project, called the Encyclopedia of DNA Elements (ENCODE), is what might be called a linguistics project for the genome. The genome at its most basic is a three billion letter text file made up of only four different letters – C, A, T and G – with no ‘punctuation’, i.e., no scientific grammar, sentences or chapters that tell us where the genes and other regulatory elements start or stop. ENCODE is determining the structural and functional elements of the genome, and is at its most fundamental an exercise in pattern determination and recognition as much as anything else.

These two large and important cataloguing projects that will tell us how to ‘read’ the genome and how each person’s ‘book’ is written a bit differently will lead to the kind of payoffs for human health that were envisioned by the initial National Research Council report that first proposed doing the HGP back in 1988 – that is, understanding of the connections between defined parts of the genome and differences in those parts among individuals, and disease incidence, natural history and response to treatment. For a variety of reasons, there has been an abundance of hype and a dearth of good data in these areas, though this is beginning to change. The mismatch between data and hype has led some to question whether genomics will really have the impact that is advertised. It’s important to remember that the genome era is less than three years old, so the field is really still in its infancy – the potential is real, but the science just hasn’t caught up with the hype yet. To me, this is a great example of Arthur C Clarke’s adage that “Everyone overestimates the impact of technology in the short term and underestimates the impact of technology in the long term.”

There are medicines out on the market now that are targeted to subsets of patients with particular genetically defined diseases, but these are mostly cancers, so target somatic, rather than germline, genetic changes. For example, Genentech’s Herceptin, which came out in the late 1990s, is an antibody against the her2/neu receptor that is amplified is subsets of breast cancers. Gleevec, made by Novartis, is an inhibitor of the bcr/abl kinase used to treat chronic myelogenous leukemia, and the epidermal growth factor receptor (EGFR) inhibitors Tarceva (Genentech) and Iressa (AstraZeneca) both appear to show differential responses depending on one’s genetic background. But the genetic differences that these drugs target are not ones that people were born with; by and large, they developed within the cancer itself.

However, we are also seeing treatments being customized based on genetic differences, or polymorphisms, that people were born with – so-called germline polymorphisms. Last year, Roche received approval for its P450 AmpliChip, which tests for polymorphisms in two common metabolic enzymes that break down drugs in the body. This is already being used to target therapies and customize doses. Since these are the first examples of genetic tests to be commonly used in practice, it is true that they are the exception not the rule in medical practice, but I have no doubt this paradigm will become common over the next 5-10 years. The best estimates are that the average drug doesn’t work in about 40 percent of the population, and we don’t yet understand how to identify the non-responders. Determining this is not only a scientific and a safety imperative but also an economic issue, since payors are increasingly demanding knowledge of whether a treatment will work in a given patient before they will reimburse for it.

It is likely that there has been such a rapid adoption of targeted therapies in treating cancer because it is fundamentally a genetic disease and it is relatively straightforward to isolate the genetic variants. Another reason is its lethality – many cancers progress rapidly, so getting the treatment right the first time is much more important and patients and doctors are highly motivated to adopt targeted therapies. There has been less motivation to target therapies for chronic diseases such as hypertension or high cholesterol as these diseases progress much more slowly, so being on a drug that does not work for a short time is unlikely to be as serious.

In the near future, promising areas for targeted treatments include cardiovascular disease and neurological diseases. Cardiovascular disease is the major killer of Western populations and a number of genetic polymorphisms have been correlated either with heart attack or stroke, or response to a variety of medicines for those conditions, so these may come into more routine clinical practice. In terms of diagnosis and understanding causation of disease, neuropsychiatric diseases (e.g., Alzheimer’s and Parkinson’s diseases) have been influenced and advanced tremendously by the advent of genetics more than any other, and it is likely that these diagnostic tests, and perhaps therapies based on them, will make their way to the clinic as well.

Translating information

Well over a 1000 human diseases have already been identified at a basic molecular level. However, simply because we understand a disease at a genetic level, it doesn’t mean we can intervene effectively to treat it. Sickle cell anemia is a classic example of this problem. It is caused by an inherited mutation in hemoglobin that was discovered in the late 1940s. It’s almost 60 years later and we have no therapy based on that molecular understanding. The challenges of translating genomic information into treatments are tough and there are systematic issues that have to be addressed to make this work.

Infectious diseases – for example, HIV and the avian flu virus – present a special problem with regard to genetics and translation, since the organisms that cause them multiply rapidly and, therefore, can mutate rapidly, presenting a genetic ‘moving target’ for therapies. This means that researchers must try to anticipate where evolution of the virus is going to go next, since these mutations can lead to drug resistance. But evolution happens so rapidly in these organisms that we’re constantly going to be in a position of having to play catch-up with these organisms in order to have targeted therapies for them.

Research universities and medical schools, government and pharmaceutical companies are all working to translate the discoveries of genomics and proteomics research into the discovery of new drugs. In the 1990s, it was, unfortunately, naively assumed that the human genome would lead to therapies quickly. This led pharmaceutical and biotech companies to dive into genomics, proteomics, metabolomics and the other ‘omics’, only to realize that the gap between genes and therapies was much larger than they thought. As a result, many companies have decided that it doesn’t make sense for them to do a lot of the early stage exploratory research looking at novel genes, and they have moved away from basic research.

Many well-known genomics companies have laid-off large parts of their genomics staffs and focused on late stage development. This is not because they aren’t interested in genomics, or don’t see its promise. Rather, it is because of the market reality that companies have to come up with a product that will generate revenue within a relatively short period of time. Unfortunately, Mother Nature is not very cooperative at revealing her secrets, and doesn’t care about quarterly earnings reports!

Of course, this leaves the question: if pharmas and biotechs are getting out, who’s getting in? One of my tasks when I moved from Merck three years ago was to help the NHGRI design and implement programs that would do exactly what we are discussing: to translate the genome into biologic and therapeutic insights. Previously, the tools for genome translation were kept proprietary, as trade secrets or protected by patents. We saw this with the advent of patents on ESTs, genes, polymorphisms, haplotypes, knockout mice, and other research tools. Virtually all of the ‘tool’ companies that were built on these sorts of technologies are now either out of business completely, or at least out of the tool business. The model now, which I firmly believe is the only way to move forward, is to put research tools and information in the public domain – i.e., make them pre-competitive. This model was established with the Human Genome Project itself, and is now being applied to a variety of other translational tool projects at NIH.

This model will, I believe, increase both the variety of projects being worked on by the private sector and the rate of discovery of new therapies. Interestingly, while the HGP has produced an enormous parts list of human function – there are estimated to be over 20,000 genes and 500,000 proteins encoded by the human genome – the pharma industry has consolidated on a number of levels, and more and more companies are working on fewer and fewer targets. If you look at the pipelines of most big pharma companies, by and large, there is a large amount of overlap, because they are all looking for validated drug targets that are seen to be relatively low risk. This is understandable, since companies appropriately need to consider their responsibilities to their shareholders and not unnecessarily take on risk. However, this also means that new drugs tend to be relatively minor improvements over their predecessors and that a new drug will have rapid follow-on competition – both patterns borne out by the recent history of new drug approvals.

My hope is that as we move from genomic information to biologic and therapeutic insights, with the tools to test new hypotheses in the public domain and freely available to all, companies will find that the risk of truly novel ‘genomic’ targets is mitigated, and that they will begin to work on them. My concern, frankly, is the time it will take for this to happen. There will potentially be a gap, perhaps up to a decade long, before this happens because the venture capital and drug development organizations have moved away from early-stage research, and the government and public sector have not yet caught up.

In terms of new technologies, systems biology and proteomics are frequently discussed, and will undoubtedly contribute a great deal. But they are in the early stages of their development and are turning out to be much more technologically complex than genomics. In the nearer term, very low cost genotyping will be a technology derived directly from genomics that will revolutionize healthcare. By genotyping I mean the ability to discover individual polymorphisms in a person very rapidly and cheaply. If the cost of sequencing continues to come down, it’s likely that by 2010 it will be possible for a person to have his or her whole genome sequenced for around US$1000. If that happens, it will fundamentally change how we think about drug development and treatment. Many of us think that our individual sequenced genomes will routinely be put on CD and taken to our GPs, so they can look for polymorphisms that would confer susceptibility to whatever the diseases are on a differential diagnosis list. There are roadblocks in the way, mainly having to do with the lack of legislative protections against discrimination in insurance and employment based on this information, but the potential is so huge for improvements to human health that I believe we will solve these problems.

Pharmaceutical and biotech companies certainly have a ‘carrot’ of potential revenue from targeted therapies leading them to work on new therapies derived from the genome. But they also have a very big ‘stick’ leading them to do so, which is the high current failure rate in drug development and the poor business fortunes many are facing as a result. This is leading companies to try new business models, including targeted therapies that may produce less revenue than a blockbuster.

Mapping our differences

The International HapMap Project is a multi-country effort to identify and catalogue genetic similarities and differences in human beings. Using the information, researchers will be able to find genes that affect health, disease, and individual responses to medications and environmental factors. All of the information generated by the project will be released into the public domain. The goal is to compare the genetic sequences of different individuals to identify chromosomal regions where genetic variants are shared.

India is rapidly gaining ground as one of the most advanced biotech hubs in the world. Krishna Deshpande, Director, Institute of Bioinformatics in Bangalore, gives NGP his thoughts on the future of genomics and why there is too much hype surrounding targeted treatments.

NGP. What have been the recent major advances in genomics and proteomics?
KD. Advances in recent years have bred a new area, Systems Biology, which aims at understanding the intricate, sequential communication and coordination between macromolecules (DNA, RNA, protein) at a global level, which will eventually take us a step closer to understanding the dynamic decisions of a cell. Systems Biology is already giving a different dimension to drug research but its success is often limited by the lack of research data in terms of macromolecular interactions. Evolution of gold standard databases providing macromolecular interaction data, coupled with systems biology will take drug discovery research significantly ahead.

NGP. Apart from sheer scale, what are the major challenges in mapping human genetic variations and how genes and proteins work together?
KD. The cost involved in genotyping stands as the major roadblock in mapping human single nucleotide polymorphisms at a comprehensive level, so as to identify genetic susceptibility to a particular disease and to the evolution of personalized medicine. Other challenges in understanding how macromolecules work together are data collection, integration and to curate the presence of false positive data, which considerably affects research and eventually results in misleading conclusions.

NGP. Some research has predicted targeted treatment solutions could be available as soon as 2010. Is this likely?
KD. I think this is a myth. No one can manufacture small molecules at will to determine what will be their side effects in a small minority of cases. I do not know of any drug without any side effects. I don’t think we are anywhere close to personalized medicine. No reasonable person can deny that there is just no way that such medicines are going to take over based just on financial considerations. In addition, we do not even have strong research data that makes discovery of such personalized drugs routine, even if one ignored the financial aspect.

There is too much hype about personalized medicine. I believe we should focus on requirements of vast multitudes of people suffering from various diseases, rather than finding drugs for targeted treatments. Though, in theory, targeted drugs in the future will certainly address the problem of differences in receptivity among the people for a generic drug. These drugs will take care of the polymorphisms which differentiate people in terms of their susceptibility to a particular disease. Additionally, the side effects due to drug administration will be brought down to the minimum.

NGP. So, overall, what are your major predictions for the future of genomics and proteomics?
KD. Huge funds are earmarked for research in two of the most important diseases: cancer and diabetes. It is not possible to find a single remedy for cancer becuase it is a complex condition. By the time you find out which proteins are involved, the condition goes to the next level and other proteins start playing a major role. As for diabetes, the chief reason for Type 2 condition is insulin resistance, developed mostly in obese people. Huge research efforts are focused on PPAR gamma platform, which is expected to reduce the size of fat cells and reduce the secretion of fatty acids.

Genomics and proteomics have already evolved into a much more comprehensive area of research, Systems Biology, and there will be rapid growth expected in the near future, primarily depending on the quality and quantity of macromolecular interaction data. This area of research is mostly collaborative in nature, involving data collection, curation, maintenance and analysis. Countries in the West, like the US and Germany, are doing a great job. Asian countries such as India are already collaborating with the West and will also play a major role in the future.