Communications - October 2011

new experiments, and help predict unobserved phenomena. More generally, the expectation is that it would allow researchers to see and understand the organism, its development, and its behavior in ways not otherwise possible.

Of course, this idea might be far too vast to be practical, but it seems worthy of consideration, if only as a very distant holy grail of sorts, toward which it would be beneficial to aspire.

Challenges for computer science. The research directions described in this article are intended, first and foremost, to yield beneficial results in biology and medicine, thus enhancing our ability to improve our lives. The central challenges they raise are also biological in nature, involving the need for biology to become a more formal, precise, and quantitative science, and the need for acquiring and consolidating sufficient information about the biology of interest to model it as a reactive system. This is especially true of the WOP and the work that is necessary to lead up to it. However, most readers of this article are computer scientists, who will be primarily interested in the new challenges this area of work raises for computer science, and in the benefits it can yield “at home.” The two are linked, of course: once our field rises to the relevant challenges, the new ideas that are found to work well in the modeling of complex biological systems will benefit the development of human-made computerized software and systems as well. So, what are the main challenges for computer science? What new ideas are needed, and what kinds of extensions should be sought for the methods used in the modeling efforts mentioned earlier?

Our feeling is that we need ways to
build models that seamlessly combine
qualitative and quantitative data, and
which come with appropriately pow-
erful analysis methods. And we need
to find ways to make our models more
robust and less sensitive to faults and
gaps in the available data. In other
words, not only biology needs to be-
come a more quantitative science, also
computer science needs to become
more quantitative. Formal methods
have excelled in structuring and han-
dling large, complex discrete systems,
but we have neglected the incorpora-
tion of quantitative data. Similarly,
we need to move our focus away from
Boolean properties of systems, such as
correctness (which really has no mean-
ing in biology), toward quantitative
properties such as fitness, robustness,
and resilience. We believe the study of
such quantitative properties will greatly
benefit computer science itself which,
as an engineering discipline, ought to
have ways of expressing and measur-
ing quantitative preferences between
different implementations of a system,
and estimating their reliability, cost,
and performance. Preliminary ideas
in this direction can be found in Cerny
et al. 13 Needless to say, by studying
biological systems in this way, we may
also learn a thing or two about building
more adaptive and robust software and
hardware systems.

acknowledgments

We would like to thank our past collaborators on these topics, for the wisdom and ideas they have contributed to us over the years. They include Luca Cardelli, Yaron Cohen, Sol Efroni, Walter Fontana, Alex Hajnal, Jane Hubbard, Na’aman Kam, Maria Mateescu, Nir Piterman, Yaki Setty, Michael Stern, Verena Wolf, and the late Amir Pnueli.

This research was supported in part by the ERC Advanced Grant LIBPR (Liberating Programming) awarded to DH, by the John von Neumann Minerva Center for the Development of Reactive Systems at the Weizmann Institute of Science, and by the ERC Advanced Grant QUAREM (Quantitative Reactive Modeling), awarded to TAH.

References

1. alur, r. and dill, d. automata for modeling real-time systems. in Proc. Int. Conf. Automata, Languages, and Programming 17 (1990), 322−335.

2. alur, r. and henzinger, t.a. reactive modules. Formal Methods in System Design 15, 7 (1999), 48.

3. amir-Kroll, h., sadot, a., cohen, i.r. and harel, d. gemcell: a generic platform for modeling multi-cellular biological systems. Theoret. Comput. Sci. 391, 3 (2008), 276−290.

4. arkin, a., ross, J. and Mcadams, h.h. stochastic kinetic analysis of developmental pathway bifurcation in phage lambda-infected escherichia coli cells. Genetics 149 (1998), 1633−1648.

5. baker, M.d., Wolanin, p.M. and stock, J.b. signal transduction in bacterial chemotaxis. Bioessays 28 (2006), 9–22.

6. barjis, J. and barjis, i. formalization of the protein production by means of petri nets. Proc. Int. Conf. on Information Intelligence Systems (1999), ieee.