By Richard Lewontin

The announcement in February 2001 that researchers had sequenced the entire human genome sparked immense publicity that was without precedent in the biological sciences. The public attention was a consequence of the  putative chief motivation in personalized medicine for the sequencing efforts. While it is true that the human genome sequence is of great interest to evolutionists trying to reconstruct human ancestry and to biologists whose attention is on the understanding the processes of development and of molecular interaction, the promise of benefits for human health has been the overwhelming justification for that immensely expensive effort.

The underlying claim is one of extreme genetic determinism. The assumption is that all-important variations in basic physiological and developmental processes are the direct result of genetic variation, so that pathological states are reflective of "abnormal" gene function due to mutations producing variant nucleotide sequences. This goes beyond the claim that systemic disorders like cancer, stroke and cardiac disease would eventually be treatable and even preventable using either gene therapy or on other interventions based on our knowledge how genetic variation causes pathological states: while the race to sequence the genome was under way, William Haseltine, as CEO and Chairman of Human Genome Sciences, assured us that "Death is a series of preventable diseases." Knowledge of our DNA apparently assures us of not only better health, but immortality as well.

The 9th chromosome, from the poster “Human Genome Landmarks: Selected Genes, Traits, and Disorders.” U.S. Department of Energy Genome Programs.


There is a long history, predating genomic studies, of the discovery of genetic mutations that are responsible for some human disorders. The classic example is phenylketonuria (PKU), in which the affected individual is homozygous for a single mutation. In PKU the enzymatic pathway that normally breaks down the amino acid phenylalanine is blocked, with the result that lethal concentrations of the amino acid accumulate in the body. The disease is rare, however, as might be expected for a simply inherited lethal disorder. We do not expect to find single gene mutations of large effect that explain the prevalence of diseases - such as cancer, stroke, and heart disease - that are the common direct causes of mortality in populations not suffering from severe malnutrition or epidemic infections. Even in the famous case of the BRCA1 and BRCA2 mutations, where the presence of the mutation results in a very high probability of contracting breast cancer, only about 15% of all breast cancer sufferers carry the mutation. The chief sources of breast cancer remain to be found.

The belief that genes determine the characteristics of individuals, together with the lack of evidence for simple single-gene defects as the cause of the major sources of disease and mortality, has led to a deterministic model of genetic causation of disease and a new approach to searching for genetic causation. This approach has been made possible by the availability of a complete DNA sequence of the human genome and of tools for detecting nucleotide differences between the genome sequences of individuals. Humans are genetically polymorphic: no two individuals (except for identical twins) will have identical nucleotide sequences. At any nucleotide position, some fraction of individuals will carry a different nucleotide than the common one, a phenomenon called Single Nucleotide Polymorphism (SNP). One technique for searching for genetic causation of diseases is to scan the genomes of a sample of both diseased and healthy individuals for positions in the genome where statistical differences exist. Ideally, all of the diseased individuals would have a different nucleotide at a particular position compared to the healthy individuals, but in  practice there is only a difference in the proportion of the four nucleotides, except in cases of well known simple genetic disorders like PKU.

The screening of whole genomes for variant SNP's has been  a major industry in both academic and commercial biomedical  laboratories during the last decade, and reports of newly discovered genetic differences between healthy and diseased individuals have been a weekly phenomenon in medical journals, in general scientific publications like Nature and Science, and in the science sections of major newspapers. Then, suddenly, it was revealed that the whole enterprise had failed to produce useful results. On the front page of the New York Times on April 18, 2009, there appeared over the byline of one of the greatest boosters of genetic determinism, Nicholas Wade, an article whose headline read "Study of Genes and Diseases at an Impasse." In the same week, the News of the Week section in the April 24 issue of Science reported on the "relatively low impact" of the SNP studies done so far. Both of these reports were instigated by an article appearing in the April 23 issue of the New England Journal of Medicine reporting on the search for "genes underlying the risk of stroke in the general population" and several commentaries on the approach to finding such genes. The general consensus of all of these reports is that the search for genetic causes underlying major causes of mortality has so far been a great disappointment.

The facts certainly bear out their pessimism. The usual measure of a specific genetic difference's importance is to calculate a risk ratio, asking: What is the risk of persons with this genotype for contracting the disorder relative to the risk in persons with a different genotype? Another form of risk calculation is the sibling risk ratio, which asks how much more likely it is that two siblings will be affected than two unrelated persons, taking into account the genes shared by siblings. In the study on stroke, two candidate SNP's were found on the same chromosome with risk ratios of 1.3 (or a 30% relative risk increase). This hardly represents a major increase in risk, but it is actually higher than the usual outcome of such studies. In a study of type 2 diabetes, for example, seven gene variants have been identified; the one with the strongest effect had a sibling relative risk of only 1.02 and the remaining six had ratios between 1.005 and 1.01.

The various commentators in the New England Journal of Medicine do not dispute these results, and one might suppose that they would begin to doubt the assumption of genetic causation - and that Wade's article in the New York Times might reflect that doubt. Yet, in actuality, their underlying assumption of genetic determination is unshaken. In reporting on the disappointing results of genome wide association studies (GWAS), Wade writes that the method "has turned out to explain surprisingly little of the genetic links to most diseases" (italics added). Moreover he states flatly that "common diseases like cancer and diabetes are caused by a set of several genetic variations in each person."

Like Wade, none of the authors of the articles in the New England Journal of Medicine has the slightest doubt that genetic differences really underlie these common diseases. What they disagree about is the best methodology for finding them. The standard GWAS method for screening for SNP's relevant to diseases is to use chips that contain about 500,000 of the approximately 3.5 billion nucleotides in the human genome sequence. These 500,000 nucleotides are those that are known to have common variants (that is, variants in frequencies of 1% or greater). The supporters of this technique point out that over 200 spots in the genome have already been shown to be associated with diseases, and as chips are improved many more such nucleotide positions will be identified. For example, 35 spots in the genome have already been associated with Crohn's disease, a form of bowel inflammation. It is these researchers' contention that the large number of variants of small effect that have been discovered  reveal the truth about the genetics of disease, namely that "many, rather than few, variant risk alleles are responsible for the majority of the inherited risk of each common disease."# They point out that new sites associated with a given disease are constantly being discovered as more samples are probed. They claim that having some information is always useful, and the fact that complete information about the genetic causes of a disease is not available at a given time does not make the method useless. However, their analysis does not make clear what preventive or curative action should be taken if scores or hundreds of individual nucleotide substitutions, each of vanishingly small effect, constitute the collective cause of common diseases.

Some of those who question this method of looking for nucleotide variants point out that the very fact that variants are in medium to high frequency is evidence that that they cannot be of major effect on the disease. Indeed, the integrated physiology of organisms makes it likely that all kinds of variation in genes whose developmental and physiological effect is far removed from the primary disease pathway will have minor effects on the disease condition. One suggested alternative method is to carry out intensive complete sequence studies of the entire genome on a small number of affected individuals. The purpose would be to find genetic variants of major effect that are at low frequency in the population and would not be detected by the chip technology. It is to be expected, after all, that genetic variants of large causal force on disease will be in very low frequency since natural selection will have been effective in reducing their frequency in the population. It is the hope of this school that a study of such rare variants of large effect will suggest therapeutic directions that are not apparent when small effect variation is studied.#

Both sides in the struggle over how to study the genetics of common disease make deep assumptions which we know not to be true. The first is their reliance upon genetic determinism.  This approach finds all diseases that are not the result of infectious agents to be the consequence of faulty genes. The failure to find the gene defects that cause a disease must therefore be the result of faulty technique; nowhere are environmental effects taken into account. At the purely methodological level, the very concept of "relative sibling risk" assumes that similarity between siblings in disease pattern must be a result of genes in common. What about the common environment within families? Is there no evidence that heart disease, cancer, and hypertension leading to stroke are induced by environmental stresses? Is it not possible that genetic effects are minor in comparison?

Secondly, there is the issue of gene-environment interaction. The methods of population sampling for genetic studies take no account of the fact that different genotypes have different sensitivities to environmental effects. Moreover, there is no reason to suppose that genetic and environmental effects are additive or even simply related. Genotype A may be more likely to lead to a disease state than genotype B in one environment, but less likely in a different environment. This is a common observation in experimental outcomes of varying genotypes and environments, yet the population sampling that is carried out for genomic disease studies takes no account of such interactions.

Thirdly, there are complex interactions between physiological and developmental pathways within organisms. Some of these interactions are of a homeostatic nature, so that the effect of large perturbations to one pathway may be dampened by reactions in peripherally related pathways, yet felt in those peripheral reactions. Just because I have a headache doesn't mean that the real problem isn't in my stomach.

The doctrine that we are the product of our DNA leads to the fantasy that by manipulating our DNA we could avoid or cure all disease and even escape eventual death. That is indeed a fantasy. All flesh is mortal.


Richard Lewontin, Ph.D., is Alexander Agassiz Research Professor at Harvard University. He is the author of numerous works on evolutionary theory and genetic determinism, including The Genetic Basis of Evolutionary Change and Biology as Ideology, The Dialectical Biologist (with Richard Levins) and Not in Our Genes (with Steven Rose and Leon Kamin).



1. Kraft, P. and D.J. Hunter 2009. Genetic risk prediction-Are we there yet?. New Engl. J. Med. 360;17: 1701-1703.

2. Goldstein, D.B. Common genetic variation and human traits. N. Engl. J. Med. 360;17:1696-1698.

Search: GeneWatch
The use of forensic DNA databases by law enforcement around the globe is expanding at a rate that should be of great concern to civil libertarians.
View Project
Other Genetic Issues
View Project