27 April 2007
Daniel Yamins
Fontana Lab, Department of Systems Biology
Harvard Medical School
Biological systems have an amazing ability to form patterns of great complexity and functionality. These feats are all the more remarkable because organisms rely on decentralized local interactions to create structure and transduce information without the benefit of a global perspective. Thinking of cells as programmable agents, computer scientists would like to find algorithms that replicate the basic powers of biological pattern formation, both to help engineer richly functional multi-agent systems and eventually to yield biological insight.
Recent work toward this goal has brought to light a surprising fact: that large classes of biological patterns can be automatically compiled. In other words, there is a general procedure that will take a desired global pattern as input and automatically design a local cellular program that will generate it. I will describe a rudimentary pattern compiler and demonstrate it in action. I will also show how it contributes to understanding the in silico analog of several biologically interesting problems, including segmentation, polarity generation, and long-range signaling.