One of the most profound revolutions in the history of science involved the discovery of the principles of quantum mechanics and the subsequent philosophical struggles to provide an interpretation of the strange probabilistic world that these discoveries revealed. The fundamental characteristic of a particle in quantum mechanics is described by its wave function. Rather than describing the position of a particle, the wave function corresponds to the probability that the particle is at any given position. According to the so-called Copenhagen interpretation of quantum mechanics (named in reference to Niels Bohr and his colleagues in Denmark), only when the position of the particle is observed does the wave function collapse to a more precisely defined location.
The peculiarity of this worldview was exemplified by a thought experiment proposed by Erwin Schrödinger in 1935. He described an experiment in which a cat is enclosed in an opaque box with a device that would release poison gas in response to a random event, such as the radioactive decay of a sample inside the box. When enough time has passed that there is a 50 percent chance that the radioactive decay has occurred, is the cat alive or dead? According to the Copenhagen interpretation, the cat exists as a superposition of a living cat and a dead cat until the box is opened and the state of the animal is observed. This interpretation led to criticism by a number of physicists, most notably Albert Einstein. However, the results of many experiments performed since Schrödinger’s proposal have confirmed the predictions of this formulation of quantum mechanics. Our world seems to be much less deterministic and more probabilistic than we intuitively imagine.
There is a potential revolution underway in medicine today. This often is referred to as personalized or precision medicine. In the spirit of full disclosure, I am now the director of the Institute of Personalized Medicine at the University of Pittsburgh. Personalized medicine is driven in large part by the sequencing of the human genome. With reference genome sequences available, new technologies have driven tremendous advances in the sequencing of individual people’s genomes. It is important to note that these next-generation sequencing methods depend on understanding the biochemistry of DNA replication. For example, pyrosequencing relies on the release of pyrophosphate associated with nucleotide incorporation into a DNA double helix as it is being synthesized. This pyrophosphate release is measured by coupling it to readily measurable phenomena such as luminescence.
With these technologies for DNA sequencing available, researchers have explored the genomic bases for many common and rare diseases. Some diseases, such as sickle cell disease, are caused by single DNA variations that are highly penetrant. For example, if an individual carries two copies of the variant hemoglobin beta chain gene associated with sickle cell disease, that person almost certainly will display clinical symptoms under appropriate circumstances. However, for many other diseases, genomic variations are not highly penetrant, but instead only increase the risk of developing the disease by a relatively modest amount. Furthermore, variations in many genes can contribute to the predisposition to a given disease.
For a number of diseases, researchers are developing models that allow the calculation of an individual’s risk for a particular disease based on these genomic variations in combination with other clinical and environmental factors. The parameters for these models generally are based on the results of studies on populations of hundreds to thousands of people. Based on the models, the risks to the individual for a range of diseases and conditions can be estimated. However, at a given time, each individual either will or will not be afflicted by a particular disease. This is analogous to Schrödinger’s cat, where the genomic and other factors are used to estimate the probabilities of an individual having a disease, but the individual either does or does not have any particular disease at a given time. Only measurements, either through the development of symptoms or through appropriate tests, result in the collapsing of the probabilistic world into a relatively definite diagnosis.
In addition to the contrast between the continuous probabilistic world characteristic of the results of population studies (that is, for example, each individual may have a 3 percent chance of having a particular disease) and the quantized world experienced by individual patients (either you do or you don’t), the revolution in personalized medicine presently is limited by tremendous gaps in our knowledge of the relationships between genotype and phenotype and between genotype and environmental factors. This is true at the diagnostic level and is even truer at the level of treatment.
This can be illustrated by comparison to another type of mechanics, namely auto mechanics. Suppose that your car won’t start when you go out to set off for work in the morning. Let’s call this condition car-won’t-start-itis. You call a mechanic, who comes out to try to start your car. If the only thing that the mechanic knows about is the role of the battery to start the car, he would perhaps check your battery or try to jump-start your car by hooking it up to his battery. If your car starts, your car-won’t-start-itis is cured, at least temporarily (although a more permanent solution may require a battery transplant). However, if this approach is not successful, he may tow your car to a repair shop for further diagnosis and treatment. This could involve testing the electrical system, the fuel pump and a range of other components of your car that could lead to the source of the car-won’t-start-itis.
This process, of course, depends on the experts’ knowledge of all of the automobile systems. Unfortunately, our knowledge of human biology and all the mechanisms by which our systems can malfunction is much more rudimentary at present. Genomic studies can reveal some of the genes whose variants can increase the likelihood of a particular disease, but the products of these genes are almost always components of one or more systems and networks. While considerable progress has been made elucidating these systems and networks, usually through undirected fundamental studies of a range of model organisms as well as studies in humans, much more remains to be clarified and discovered. This incomplete knowledge presents great opportunities for biochemistry and molecular biology as well as younger fields such as computational and systems biology.
In 1929, Paul Dirac, one of the primary developers of quantum mechanics, noted that “the fundamental laws necessary for the mathematical treatment of a large part of physics and all of chemistry are known … and the difficulty lies only in the fact that the application of these laws leads to equations that are too complex to be solved.” Of course, many scientists have worked diligently in physics and chemistry over the past 80 years developing methods to solve these equations approximately and to generate empirical data to be analyzed to reveal principles that are not generated directly from theory alone.
I expect that personalized medicine will follow an analogous path, with some insights available directly from analysis of the genome sequence but with most substantial progress dependent on the challenging integration of empirical and more mechanistic and fundamental information. In addition to the tremendous opportunities for biochemistry and molecular biology and ancillary fields noted above, we, both individually and as a society at large, will have to come to terms with the world of probabilities that is emerging before us. We all will have to become much more conversant with distinctions between relative risk and absolute risk. A certain genetic variant may be associated with an increase of 50 percent in your relative risk for developing a particular disease. This can sound quite worrisome. But if the average absolute risk of developing the disease is only 1 percent, then your absolute risk increases to only 1.5 percent. Such factors represent components of our own personal wave functions, and we will all have learn to interpret these wave functions and to deal effectively with the results when these wave functions collapse into measurable reality.
Jeremy Berg (firstname.lastname@example.org) is the associate senior vice-chancellor for science strategy and planning in the health sciences and a professor in the computational and systems biology department at the University of Pittsburgh.