Let us be clear lest the above suggest that we are pessimistic about this field. There are very few disciplines within biomedical research with as much promise and excitment as functional genomics. Consider the example of the analysis of large B-cell lymphoma, a deadly malignancy of the lymphatic system, conducted by Alizadeh et al. . In (a necessarily abbreviated) summary of their study, the gene expression of the lymphatic tissues of a cohort of patients with the diagnosis of large B-cell lymphoma was measured using DNA microarray technology. That is, thousands of genes expressed in these tissues were simultaneously measured in the respective lymphatic tissue sample of each patient.
When a clustering analysis was performed to see which patients ressembled one another the most, based on their gene expression pattern, two distinct clusters of patients were found. When the investigators examined the patients' histories it became apparent that the two clusters corresponded to two populations of patients with dramatically different mortality rates (illustrated by the survival curves in figure 4.13, section 4.11).
The implications of these two significantly distinct mortality rates are profound. First, these investigators have discovered with genomic technologies and bioinformatics analyses a new subcategory of large B-cell lymphoma, a new diagnosis with clinical significance. Second, they have generated a tool that provides (pending confirmation in other studies) a new prognosis; patients can be given much more precise estimates of their longevity. Third, it provides a new therapeutic opportunity; patients with an expression pattern predicting a poor response to standard therapy may be treated with different (e.g., much more aggressive) chemotherapy. Fourth, it presents a new biomedical research opportunity; what is it about these two subpopulations that makes them so different in outcome and how can that be related to the differences in gene expression?
It is rare that a set of measurement and analytic techniques can so revolutionize biomedical research and clinical practice. It is precisely because the excitement and expectations surrounding this field are so high that we are compelled to inject a note of skepticism about the measurement techniques and the analytic methods. Without such skepticism, the scientific method cannot function and the field will not advance. Nevertheless, even as we have discovered for ourselves significant drawbacks in genomic methodologies, which we address in this book, we remain convinced that these problems only represent the transient growing pains of a new field of biomedical investigation.
Who is this book intended for? Answering this question has served as our constant compass throughout the book's writing. There are three audiences that we have had in mind.
1. Experienced biologists with limited experience using expression microarrays, or who are concerned that their current approaches to this field are problematic. For them, this book provides a systematic approach to using microarrays as a tool to investigate biology. Our particular goal for this audience is to realize the pitfalls and caveats relating to expression microarray technologies and their analysis. Also, we intend this book to provide these biologists with a firm foundation on which they can engage in collaborations with colleagues formally trained in bioinformatics and biostatistics. We deliberately limited the amount of mathematical treatment of this material. Where it is present, the reader can skip it without significantly reducing the comprehension of the subsequent text.
2. Experienced informaticians with limited experience analyzing microarray data. We believe the field of functional genomics to be one of the most fruitful and rewarding for a computer scientist or other quantitatively trained scientist to enter. It provides a source of challenges, problems, and data sets that will stimulate basic methodological development while furthering important goals in the enterprise of biological discovery and the state of the art of clinical care. For this reason, we emphasize an approach that is driven by the questions of interest to these latter goals. As a result, this book will not present the fundamental computer science underlying various techniques (e.g., proofs of the soundness of various machine-learning algorithms) but will instead illustrate their application to problems that are challenging investigators in functional genomics. This is not to say that the approach we have taken is not rigorous; it just eschews details on the methodologies that are available elsewhere and that we have copiously referenced.
We are convinced that the most productive research projects will be those that involve intimate collaborations with biologists. This is in contrast to the remote and post hoc analysis of the outcome of experiments that the bioinformatician has little to do with, but which is frequently the norm in this discipline. We intend and hope this book can serve as the basis for collaborations in which the informatician understands the goals of biology in this scientific enterprise and in which she or he can contribute to the experimental design and analysis as a first-class member of the research team.
3. Students entering the field of Bioinformatics. In our own classes (e.g., Medical Computing 6.872 taught at MIT), we have been gratified to note the emergence of a new generation of students who are both facile in the use of computers as experimental tools and who have a broad understanding of the biological sciences. Although this text can be used as the basis of a course, it is also designed for independent study. For those students, this text is intended to serve as a rapid introduction to the fields of microarray-driven studies of functional genomics so that they can become productive researchers even while they pursue their studies. Indeed, we have had several successful collaborations with undergraduates who have used this material as a launching pad for graduate research or careers in industry.
Was this article helpful?