Methods for Combining Experimental Data and Molecular Simulations into Hybrid Models

Creating models that are consistent with experimental data is an essential task in computational modeling. This is generally done by iteratively tuning the input parameters of a simulation to match experimental data. An alternative method is to bias a simulation to match experimental data leading to a hybrid model composed of the original model and biasing terms derived from experimental data. Recently, we've developed two methods that can be mathematically shown to bias a simulation as little as possible to match experimental data. One method is for matching PMFs in collective variables space (e.g., RDF) and another for ensemble averages. I'll report progress on biasing dynamical quantities as well and present results on applying these methods in both ab initio models and molecular dynamics simulations.