April 26, 2012 - Rebecca Schulman

Title: The Computational Power of Chemical Reaction Networks

Abstract:
Single living cells can chase targets, change shape in response to
functional cues, and self-replicate.  How can a simple group of
molecules inside a cell orchestrate such complex behaviors?  While
individual molecules may have limited function, a network of chemical
reactions can simulate a Turing machine, and it is widely believed
that the complex control algorithms in cells are the result of
networks of interactions involving many molecules.  I'll describe how
we can investigate the power of such networks of molecular
interactions by trying to reimplement computations, construction
processes and control algorithms as molecular interactions involving
synthetic DNA molecules.  The chemistry and structure of DNA is
well-understood, and we can engineer specific interactions between DNA
molecules by designing their sequences.  We can therefore focus on the
dynamics of systems of interactions rather than the chemistry of
individual interactions.  I'll show how we can use DNA to build a
self-replicator whose alphabet is a series of DNA blocks and program a
set of molecules to execute a "search and capture" process that can
form tether between two points of unknown location.  From these
examples we learn that molecular reaction networks are surprisingly
powerful: a relatively small set of molecules can both compute and
learn arbitrarily complex patterns, and even though molecular
interactions are stochastic and unreliable, we can design systems of
molecules whose behavior is robust.  I'll close by describing some new
work designing computational "morphogenesis" processes that could
allow us to produce complex standing chemical patterns in 3
dimensions.

Bio:
Rebecca Schulman studied computer science and mathematics at MIT,
where she worked with Gerald Sussman's group as part of the amorphous
computing project.  After a several year hiatus in Silicon Valley in
which she helped start a company focused on natural language access to
databases and where she wrote Linux software at Eazel, Dr. Schulman
returned to graduate school to study molecular computation.  She
received her PhD in computation and neural systems from Caltech in
2007, where she studied under Erik Winfree.  From 2008 to 2011 she was
a Miller research fellow in the physics department at the University
of California Berkeley.  Dr. Schulman is currently an assistant
professor in chemical and biomolecular engineering at Johns Hopkins;
her group focuses on the design and characterization of complex
self-assembly processes and the construction computational materials
and structures.