As computational biology and systems biology become mature disciplines, they need the kind of research attempted here. Many computer scientists have built complex, robust, sophisticated tools to model and analyze biological systems; over the last dozen years, several of these models have been used to develop accurate and detailed models of many well-known biological systems, including several metabolic and signaling gene networks. It can be argued that many of the resulting publications show how computational models reflect and confirm known understanding, but most tools have yet to bring about a better, deeper understanding of some less-understood systems, or to offer truly novel insights and suggestions for experiments to confirm them.
In this paper, the authors make use of the bio model analyzer (BMA) tool, developed primarily at Microsoft Research. The BMA uses Silverlight to provide a web-based graphical interface where networks of proteins can be drawn and placed within spaces identified as cells, and numerical values can be assigned to the concentrations and reaction parameters. Once a system is built, the user invokes the proof mode, with which semantic consequences of the system can be teased out. The process emphasizes the proof of stability, to the extent that each species of protein eventually reaches a fixed concentration value in every execution.
The introduction to the BMA tool is not particularly clear, nor is the description of how to use model checking when reasoning about genetic networks. Instead, the paper focuses on the issue of the usefulness (as opposed to the usability) of the BMA and of computational tools for science in general. This is done by walking through a case study with a biologist named Lucy (who is curiously not one of the authors of the paper), who expresses her reactions while using the BMA tool to study the genetic network underlying blood cell development. The description of her motivation points out one of the main potential benefits of modeling: being able to introduce perturbations into the system and prompt the shaping of “new hypotheses and ultimately new understandings.” The difficulties Lucy has with the particular tool being used (BMA) are described, including finding the right level of abstraction and understanding the subtle notion of time used in proofs obtained by model checking (as opposed to the more direct notion of time used in simulation tools).
The description of this case is interspersed with more general discussions about using computational models as scientific tools. A contrast is made between a bottom-up approach, where experimental data drives the construction of a model, which simulates a real entity, and a top-down approach, such as that underlying BMA. The explanation of this top-down approach is somewhat fuzzy, which is unfortunate because it is an important aspect of the discussion. The authors state that the model is derived from the execution of a proof regarding the stability of the system. The initial model is then presumably refined with more details, and each refinement is only accepted if it maintains the desired property of stability.
It would have been useful to have a concrete example of this process of initial model postulation and refinement applied to a gene or protein network, but unfortunately none was included. Lucy’s work is only explained at a high level, and apparently none of the screen images are actually from this case study.
I also found the paper confusing because it mixes issues involving the use of BMA with more general, philosophical questions. For example, would any tool that intends to encourage users like Lucy to use a model checker have these same issues? How do bottom-up simulation tools for gene networks fare in terms of “modes of knowing”?
The conclusions in this paper are very interesting for anyone working in computational and systems biology, but they barely begin to address the issues. Maybe this paper should be seen as a starting point for a fruitful discussion on the role of computational techniques in biology--not in some idealized setup as imagined by tool builders, but in careful examination of the mental and philosophical underpinnings of biologists who are immersed in the details of problems and who may be willing to invest in using tools that can truly provide a significant new way of obtaining valuable insights.
