Figure 9
From: Unbiased, scalable sampling of protein loop conformations from probabilistic priors
Free-endpoint chain sampling with heterogeneous priors. Comparing SLIKMC against a standard Metropolis-Hastings (M-H) sampler on a free-endpoint chain with heterogeneous prior distribution over joint positions (crosses: means, shaded circles: 3σ spreads). For each method, 400 iterations are run and every 20'th sample is retained, taking approximately 10 s time on a standard PC. Left: M-H takes steps that are too large and only generates 4 unique samples. Middle: M-H with step size reduced by 10 has a higher success rate but slower convergence. Note the lack of variance in the leftmost point. Right: our method.