Opinion- OR & Computational Biology: A Career Opportunity
Harvey J. Greenberg
University of Colorado Denver
hjgreenberg@gmail.com
Allen G. Holder
Rose-Hulman Institute of Technology
holder@rose-hulman.edu
November 3, 2011
The life sciences have undergone sweeping changes over the last couple of decades, and modern biological studies require elements of physics, chemistry, engineering, mathematics, and computer science. These host disciplines mirror the origins of operations research, and as the overlap suggests, there is a growing symbiotic relationship between OR and the biological sciences. The focus of this article is to briefly discuss this relationship and to highlight a promising future that OR and computational biology share.
One of OR's signature strengths is taking an inter-disciplinary approach to problem solving within a team, and since most of the research in computational biology is accomplished in collaboration with relevant specialists, an operations researcher finds him or herself in a native research environment. The difference is that unlike traditional OR applications in business and engineering, biological problems use a different language. However, an operations researcher is accustomed to learning enough about an application to obtain an understanding of a problem from its source. Once armed with a firm understanding of the biological question, our training in computation and mathematics makes us especially well suited for a discourse across the research team. Hence, an operations researcher can be integral to a team's cohesion.
An operations researcher has several technical skills that are valued in computational biology. The field emerged from the sequencing of the human genome, and today's research questions are rich and broad in scope, ranging from increasing crop yields to targeted, DNA-specific drug design. OR methods can have a bilateral relation with the scientist's laboratory experiments. Experiments provide some data to guide the development of a computer-resident model and applying it for analysis, which in turn guides the next experimental designs. Experiments are often extremely costly compared to computer simulations. Computational research can be used to answer investigative or analytical questions. In the former we typically have a computational model of a biological system; in the latter we often have copious amounts of experimental data. Much of the research is naturally and intuitively modeled in the language of optimization, stochastic processes, and graph theory, and in many instances these OR stalwarts have become the workhorses of the research. From an OR perspective the applications may have changed from management and engineering to biology, but the fundamental research in modeling and computation that has drawn many of us into OR remains the same. Similar to OR's history in management and engineering, both OR and the host discipline gain: OR grows through the study of new and adapted models and solvers, and computational biology grows through new science.
Computational biology particularly welcomes OR's expertise in modeling and algorithms. Model formulations of biological phenomena are not unique, and an operations researcher is particularly astute at knowing the strengths and weaknesses of various models along with the pragmatics of solving their instances. We also know the important details of using different software components and the subtlety of linking models with a suite of solvers. Our training in the design and analysis of algorithms further complements our modeling expertise. In short, our modeling and algorithmic prowess can be a research advantage over competing groups who instead use 'black-box' software without an advanced awareness of the interplay between computed solutions of model instances and relations to the actual problem at hand.
To highlight one area of research, consider systems biology, which is concerned with large and complex models of the interactions among biological components. One such computational study investigates the metabolism of an entire cell. Suppose we seek to maximize the production of some metabolite in a network composed of biochemical reactions that consume some compounds and produce others. A modeling approach is flux balance analysis, where equations describe an input-output model just like a Leontief model of a closed economy. The model is a linear program. In OR we often turn the question around, “Why is this pathway different?” We know to focus on extreme pathways and obtain signatures: For what objective is this pathway uniquely optimal? This enables us to interpret the biochemical pathways and differentiate among organisms. We also know how to use binary variables to represent logical conditions, such as blocking some pathway while minimizing the effects on others. Thus, mixed-integer programming does the trick. Now add some uncertainty. A metabolic network has regulatory controls in which a gene produces a protein that can inhibit or activate some reaction. This occurs with some probability, and one model is that of a Markov process. Also, certain cells go through a process that has the stochastic properties of a queueing system, with the outcome of the process being cell death a portion of the time. Not dying frequently enough could cause tumor growth, which we would like to prevent. Simulating this process permits us to study the likelihood of such an unwanted event.
For those entering their professional lives in operations research the opportunities in computational biology are substantial. Federal funding agencies include NIH, DOE, and NSF, and there are non-government foundations plus a variety of industries, notably pharmaceuticals, that all support research and post-graduate studies, sometimes jointly. There is also an increasing number of journals for OR and computational biology, notably the INFORMS Journal on Computing. Most disciplines have communities within them that share a common interest in OR and computational biology, and of course we have the INFORMS Computing Society. These communities organize conferences and sessions within the parent society to stimulate collaboration and promote research.
In our opinion, the career opportunities in computational biology are only just emerging. OR offers a great launching platform to succeed in this burgeoning discipline, and thanks to many primers, survey articles, and texts that have been written during the past decade, the task of getting involved is becoming easier. The authors encourage participation in this exciting frontier.