The March issue of Scientific American features an article by Francis Collins, head of the National Human Genome Research Institute and Anna Barker, deputy director of the US National Cancer Institute, setting out how to wage a smarter War on Cancer (1). This latest initiative, packaged as The Cancer Genome Atlas (or TCGA for cute), aims to catalogue all the somatic mutations in selected lung, brain and ovarian cancers as a $100 million pilot project and then anticipates moving on to catalogue mutations in another 50 or so cancers, provided further billions of dollars of taxpayers money is forthcoming. This initiative was originally highlighted as the Cancer Genome Project in the New York Times (2) and in Nature Biotechnology (3), and is the brainchild of Eric Lander of the Broad Institute (4). When in full production, TCGA anticipates sequencing the DNA of over 12,000 tumor samples to reveal all the DNA mutations within them, the underlying assumption being that the generation of tera bytes of cancer DNA sequence data will seamlessly equate with drug-based therapeutic empowerment. The reasoning goes as follows. By knowing all the mutations in a primary tumor sample from a patient, a physician will be able to prescribe a cocktail of drugs together with a chemotherapeutic regimen which is specific for that individual. Any new drugs will be intelligently designed by pharmaceutical companies to match the alterations in the defective proteins that arise from these DNA mutations in primary and metastatic tumors. In a nutshell, this is the Holy Grail of personalized cancer medicine.
In their article, Collins and Barker chart the challenges that lie ahead by using the analogy of the original 1804-1806 Lewis and Clark expedition to explore the unknown Northwest territories on orders from President Jefferson. Collins and Barker seek gravitas and the high ground by highlighting the input of Nobel Laureates who look favorably upon their endeavor. However, reputations count for little in any war and the War on Cancer is no exception. This one is long running, shows little sign of being won and has continually raised and then dashed the hopes of cancer patients. By contrasting the practicalities of the bedside with the pure research efforts, we illustrate how TCGA could benefit from more cerebral input prior to such profligacy of DNA sequencing. Readers can judge for themselves whether the strategy put forward in the Scientific American article is indeed as momentous as claimed, or whether this is another Roadmap within the Land of Unfulfilled Promise.
The knub of the problem is that tumors are heterogeneous collections of cells which together contain millions of genomic perturbations all the way from aberrantly methylated bases to gross genomic imbalances at the chromosomal level. The central issue of clinical genetics and of cancer has always been the following. How do we sort the mutations of very different types which may produce a clinical outcome in the unique genetic and epigenetic background of that particular individual, from those flotsam and jetsam mutations which are clinically asymptomatic. ...
Most importantly, however, only 1 in 50,000 or so of the cells in a primary tumor ever develops the requisite genomic alterations to emigrate and embark upon a journey which will leave its descendants in far off anatomical locations. It is these rare rogue cells that pose the danger, not the cells that make up the bulk of the primary tumor itself. Thus analyzing an entire solid tumor for all its mutations, as codified by the NCI and the NHGRI, is close to an exercise in futility. ...
A far more important finding from these data, which is not emphasized either by the above authors or by Collins and Barker, is that the mutational signature of each patient is unique (14), a finding which has enormous implications for drug treatment. ...
If TCGA does indeed have therapeutic utility, it should be a straight forward matter to match the mutational spectra generated by the Vogelstein, Kinzler and Velculescu laboratories with an already approved FDA drug combination for each patient. ...
The fact that robust therapeutic predictions have not emerged from the mutational data is a telling indictment of TCGA strategy. TCGA is certainly keeping sequencing centers busy, but offering little in the way of therapeutic clarity. ...
The take home message is that the cells which leave a primary tumor consist of a heterogeneous population with varying combinations of translocations, inversions, chromosomal as well as extrachromosomal amplifications, segmental aneuploidy and assuredly, aberrant genomic methylation. The rules by which somatic mutations can exert their influences on phenotype, when they are in imbalanced genomes, are radically different from those that apply in diploid oness. ...
The fact that most metastatic cancers are almost as deadly today as they were 50 years ago, is information that is publically available from the NCIs own databases and examples are illustrated in Fortune and Nature Biotechnology (7,3). We reiterate that it is when those rare genomically heterogeneous cells in a solid tumor begin to emigrate that the serious problems really begin. Dissemination can occur very early when tumors are only millimeters in diameter and probably even before this size is reached (18). The clinical data of Tarin, Riethmuller and Klein, as well as the follow up data from more than 300,000 cancer patients in the Munich Cancer Registry gathered over a 20 year period, unambiguously demonstrate that the majority of cells in a tumor never acquire the ability to either disseminate or to metastasize; this property is only achieved by the rare so-called stem cells now being described in a plethora of cancers. From the vantage point of the surgeon and the pathologist, it is indeed axiomatic that if the majority of cells in a primary tumor could metastasize, as inferred from microarray–based studies (25), there would be no residual tumor in its original location (10).
The clinical and so-called cancer stem cell data clearly expose the danger of relying on bioinformatic analyses, as Lander and colleagues have done, to extrapolate from microarray data to clinical importance (25). Even though their primary and metastatic tumor samples did not even come from the same patients, these authors dismissed the notion that important signatures arose from rare cells within the primary tumor and instead placed their faith in signatures from whole tumors. However, cancer signatures from whole tumors are notoriously fragile; a fact which has been well documented by leading statisticians and mathematicians such as Tibshirani and Domany (26, 27). By contrast, clinical and pathological observations on patients and insights from the so-called cancer stem cell compartment, are far more informative than the artefact-prone results from expression-based analyses of whole tumors. Readers will now better appreciate that the therapeutic foundation of The Cancer Genome Atlas was predicated on a false premise; that the mutational signature of the bulk of the primary tumor would be congruent with its metastatic derivatives. The perplexing issue is that primary and metastatic tumors were already known to differ markedly at the gross genomic level years before the TCGA was proposed. Therefore its proponents are either still oblivious to a large amount of classical genetic, cell biological and clinical data which impinge on somatic processes, or choose to ignore them. The patient data which show that primary tumors and disseminated cells have different signatures at the single cell level (17, 18), reduce the TCGA to a weak force. Prescriptions of personalized drug combinations based on signatures from the bulk of a primary tumor are an illusion, whereas those from the so-called stem cell population still offer some hope.
In a therapeutic context, the description by Collins of The Cancer Genome Atlas as a dream come true and the call to arms by Collins and Barker of imploring scientists to think outside the box, is fatuous. We shall be much more enthusiastic about encouraging thinking outside the box, when there is concrete evidence of any thinking going on inside it. [Read the entire article.]
The idea of there actually being a contemporary war on cancer is a farce. The prevailing paradigm of genomics has prevented any progress in this area of research. In fact, genomics can be blamed for many of the faults in modern-day science. This nasty field of study pervades all areas of academia, and brings with it a very sad conclusion – that correlation is the almighty statistic.
I find genomics’ influence saddening, and will never understand why researchers in this area are not expected to prove causation via the same mechanisms as infectious disease researchers. Although, I guess infectious disease researchers are not so rigorous anymore either (sigh).
Posted by: An academic youth | March 06, 2007 at 10:24 AM
Academic youth,
Below is a taste of what the molecular biology that was, was like. Today's version bears no relation, as you so accurately and so sadly note.
"In molecular biology, structure and function go together like “love and marriage” in the song. And in the molecular biology in which Peter honed his scientific acumen—one without the option of material gratification—the only song anybody was singing was some version of “my brain’s bigger than yours.” This meant that no one would dare present as anything more than a working hypothesis a conclusion based on an incomplete demonstration that a particular mutation was responsible for a difference in function.
Ultimately this would mean having an x-ray crystallographic analysis of the protein (the gene in question’s product) in both the normal and mutant forms in which the particular difference (attributable to the mutation in the gene) is not only evident in an altered three-dimensional structure, but in which the altered form can be directly related to differences in the functional properties of the protein. It’s a tall order, and one that is rarely filled. It is (or was) known as a complete genetic analysis, and accomplishing it was the goal of every molecular biologist worthy of the name. Peter’s and my friend Alvin John Clark, who began teaching me bacterial genetics in 1966, has had a marvelously productive life in science doing just that. He completed a single genetic analysis, beginning in 1965 with the isolation of a mutant bacteria unable to recombine its DNA properly, and finishing almost thirty years later with completely satisfying explanations, from the biological to the atomic levels, of why this is, how it happened, and what it means.
But Peter was by no means presenting such a steep challenge in the review article. Short of the proof that comes with crystals, there are many steps that can be taken to show that the products of two genes are likely to have similar functions based on analysis of only the gene parts of the gene-protein totality."
Oncogenes, Aneuploidy and AIDS: A Scientific Life & Times of Peter H. Duesberg, pp. 11-12.
Posted by: Harvey Bialy | March 06, 2007 at 10:51 AM
Dr. Bialy
The time of science has come and gone. I can only hope that it will return for the enjoyment of the few of my generation, who I know can actually participate in the thought process.
The majority are surrounded by technology, which they do not understand. Results are garnered and reported only if they look right. Those who actually understand the technology are a minority.
Sociology may be called a soft or pseudoscience, but a sociological understanding of the politics of the hard sciences is quite damning.
Posted by: An academic youth | March 06, 2007 at 11:21 AM