The rational design of small molecules for use as potential
therapeutics or chemical probes for understanding biology is
immensely difficult. Perhaps surprisingly, engineering novel
small molecules that selectively bind biomolecular targets
with high affinity, are soluble enough to reach useful
physiological concentrations, and can permeate cells can all
be expressed as ratios of normalizing constants (partition
functions in statistical mechanics) that can be computed using
Markov chain Monte Carlo techniques. We will give an overview
of recent developments in efficient MCMC methods that borrow
from both statistical inference and nonequilibrium statistical
mechanics to greatly accelerate the rate at which these
quantities can be computed, and describe new algorithms that
have the potential to automatically and efficiently guide the
design of novel small molecules.