If microarray technology is going to be commoditized such that it can be used routinely as a measurement system, even in relatively unskilled hands, then what will distinguish that particular genomic enterprise with aspirations after excellence?
The most pressing need and most obvious quality that will distinguish the best functional genomics investigations from the rest will be the nature of the investigators participating. As to which investigators these might be, the genomic pipeline outlined in section 1.5 points to the kinds of expertise that will be most in demand. There will need to be top-flight basic biologists, pathologists, clinicians with access to populations of interest, all capable of and motivated toward posing the high-yield clinical questions that are likely to drive many of these investigations. Based on the experience of several of the more successful laboratories, it seems that at least for the near future, the largest and most pressing need for expertise is in the area of bioinformatics. A typical successful laboratory will have approximately two thirds of its investigators drawn from the ranks of bioinformaticians. This proportion is likely to be maintained for at least the next 5 to 10 years, until the analytic tools are much more robust and the analytic approach is systematized and standardized. At that juncture, it may be that bioinformatic analyses will be packageable in the same way as common biostatistical analyses for clinical trials are packaged in programs such as SAS (Statistical Analysis System) or SPSS (Statistical Package for the Social Sciences). Of course, the availability of such packages allows neophytes to perform uninformed, flawed, but superficially good-looking analyses, a problem that most biostatisticians are all too aware.
For the foreseeable future then, the most significant investments in expertise will probably have to be made in the area of bioinformatics. As individuals with training in biology and computer science and and medicine are rare and hard to come by, most laboratories are currently hiring individuals with strong computational skills, such as computational physicists, mathematicians, and statisticians, and providing them with the intellectual tools and background to become bioinformaticians. This process is not without its problems and uneven results.
A more lasting and productive process is to develop training programs that will produce the necessary interdisciplinary education in biology, computer science, statistics and probability theory, to mention just a few desirable knowledge bases that those individuals should be equipped with. The National Institutes of Health have recognized this need and are beginning to fund several training programs. Locally, the Division of Health, Sciences, and Technology (HST) at Harvard University and the Massachusetts Institute of Technology has developed a Bioinformatics and Integrative Genomics (BIG) training program. BIG is designed to take students with strong, quantitative undergraduate backgrounds such as chemical engineering, mechanical engineering, physics, mathematics, or computer science and then provide them with a curriculum that involves clinical exposure at Harvard Medical School during clinical rotations, joint classes with Harvard Medical School students, as well as basic biology courses. This and advanced topics in probability theory, computer science, and other genomics-related courses will provide these students with the basis to complete their doctoral thesis at HST in genomic science and subsequently to assume leadership roles in genomic research.
Even now, when many of the genomic measurement techniques have in fact not yet been commoditized, it is apparent that the most expensive and labor-intensive part of the genomic enterprise will be the accurate and large-scale phenotyping of human populations and model organisms. Specifically, it takes a lot more time and personnel to accurately document all the characteristics of a human pedigree with all the attendant clinical findings and measurements of each member of the pedigree, than to simply genotype or obtain expression measurements on each member of the pedigree. Consequently, any substantive forward-looking genomic effort will involve a carefully thought-out plan for the acquisition of detailed phenotypic information for all obtained samples. This activity has previously been most associated with genetic epidemiologists who have carefully obtained prospective perspectively gathered phenotypic data in well-designed studies such as the Nurses' Health Study, the Physicians' Health Study, and the Framingham Heart Study.
With motivations similar to those of epidemiologists, prospective annotation and phenotyping are indeed what are required. Those who expect to be able to mine the existing clinical records of a patient's care for the necessary data are going to be disappointed. Even if they can manage the issues of patient consent and data release, the quality of the clinical record and its emphasis is such that most of the data needed for effective and reliable phenotyping of the clinical population are unavailable. Consequently, any serious attempt to achieve high-throughput phenotyping for a substantive study will require that a laboratory or institution invest in the necessary infrastructure for such activities. These activities are labor-intensive and therefore not inexpensive. Specifically, it requires the use of trained personnel, such as genetic counselors, physician assistants, or highly trained research assistants, who are aware of the complexity of human phenotyping. As human subjects have relatively long lives as compared to microbial or murine models, these activities have to be supported adequately over decades. High-throughput human phenotyping also requires that the institutional laboratory put in place the consent and data release procedures that will enable the research to be done subsequently, and to ensure that these procedures are consistent with the aforementioned rules and regulations.
The importance of prospective phenotyping activity cannot be overemphasized. Without it, we will never be able to bridge the gap between the observed and measured phenotypic and genomic data and the clinical import of these data. This is a far less sexy activity than other aspects of the functional genomic enterprise, but it ultimately will be what distinguishes those efforts that are the most successful and clinically relevant. Who best performs such prospective high-throughput phenotyping will determine whether the locus of human-related genomics will be under the administrative ¿gis of hospitals, medical schools, the industrial sector, or governmental organizations.
Was this article helpful?