An information-theoretic approach to improve the estimation of rare event probabilities in biochemical systems

The stochastic simulation algorithm (SSA) is extensively used to simulate biochemical systems. By generating realizations from the probability distribution governing the time evolution of each system, the SSA provides a numerical approach for quantitative estimation of system behavior. A consequence of this approach is the dependence of estimation accuracy on the number of realizations generated— higher accuracy requires larger sample sizes. This requirement can render rare event probability estimation intractable, as the number of realizations needed to witness even a single rare event can be very large.