On the use of experimental observations to bias simulated ensembles

Historically, experimental measurements have been used to bias biomolecular simulations toward structures compatible with those observations via the addition of ad hoc restraint terms. We show how the maximum entropy formalism provides a principled approach to enforce concordance with these measurements in a minimally biased way, yielding restraints that are linear functions of the target observables and specifying a straightforward scheme to determine the biasing weights. These restraints are compared with instantaneous and ensemble-averaged harmonic restraint schemes, illustrating their similarities and limitations. © 2012 American Chemical Society.