Scientific Computing in Molecular
Biology
Frank Noe,
Free University of Berlin
Abstract: The application field of molecular biology has long resisted
mathematics and in particular classical scientific computing due to a
number of key problems: (1) extremely high dimensionality, (2)
extremely large span of relevant timescales, and (3) inherent
stochasticity. About 10 years ago, the transfer operator approach
has been introduced which has allowed to reformulate molecular dynamics
in a way that provides (1) a relatively low effective dimensionality,
(2) only maintains the (interesting) slow timescales, and (3) describes
the deterministic transport properties of the system. This approach
allows scientific computing approaches such as error control and
adaptive refinement to be used to solve the molecular kinetics even in
cases where a direct simulation would be computationally unfeasible. In
this talk I will describe the transfer operator technique and
illustrate its usefulness to modeling the kinetics of complex molecular
systems and understanding the relation between microscopic
(computational) and macroscopic (experimental) observables.