Exploring similarities of network based information processing in biology and computer systems
Networks are the backbone for information processing in distributed systems. Such systems are prevalent
in computer science and biology, both of which rely on networks of interacting entities to reach joint
decisions, coordinate and respond to inputs. There are many similarities in the goals and strategies of
biological and computational systems which suggest that each can learn from the other. These include the
distributed nature of the networks (in biology molecules, cells, or organisms often operate without central
control), the ability to successfully handle failures and attacks on a subset of the nodes, modularity and
the ability to reuse certain components or sub-networks in multiple, and sometimes very different,
application and the use of stochasticity in biology and randomized algorithms in computer science.
These observations, some dating back to the 60’s, have inspired the development of several computational
methods including neural networks and genetic algorithms. However, this early work relied on a highlevel
(and often flawed) understanding of biological processes. While these methods led to several
successful applications, they usually did not lead to new insights regarding the ways in which information
is processed in biological networks. Similarly, little thought has been given to how such information
processing in biology can serve as a basis for robust and adaptable methods for computer networks.
Over the last few years we have began to witness a change in how biologically inspired computational
methods are derived and studied (in part due to the large increase in our ability to collect and analyze
biological data). A number of recent bi-directional studies, by us and others, have demonstrated that
thinking computationally about the settings, requirements and goals of information processing in
biological networks can both, improve our understanding of the underlying biology and lead to the
development of novel computational methods providing solutions to decades old problems.
In this proposal I discuss a set of studies that, combined, will provide a coherent understanding of how
robust and adaptive computation and coordination can be performed over networks even when resources
are severely limited, communication is minimal and noise and attacks are prevalent. Specifically, I will
attempt to answer the following three questions: 1. How can we design robust and efficient networks? 2.
What type of communication is required to compute and coordinate over these networks? 3. How can we
propagate information over network nodes even with limited connectivity and communication? Each of
these questions will be addressed by coupling a specific biological process (in yeast, fly and E. coli.) with
related computational problems (ranging from security to distributed computing to machine learning).
Answers to these questions will shed new light on the biological process and how it uses networks to
achieve its goals. In addition, they would provide new directions for addressing the relevant
computational problem. Our overall goal is to highlight the similarities between network based activities
in these two domains and demonstrate that thinking about them from an information processing point of
view can improve our understanding of systems in both domains.