The effect of different methods for incorporating experimental data on a simple example consisting of a mixture of two bivariate normal distributions.

<p>In this example, we only have experimental data regarding the y-dimension of the distribution (target value indicated by dotted line). The top row contains the unperturbed and maximum entropy distributions. The matrix shows various combinations of force constant () and number of replicas () when enforcing the restraint through a harmonic potential. In these calculations, corresponds to the standard method for structure calculation, and corresponds to ensemble refinement. In each plot we also show the mean in the y-direction (), and the entropy of the distribution ().