MELD-Adapt: On-the-Fly\nBelief Updating in Integrative\nMolecular Dynamics

Integrative structural biology synergizes experimental\ndata with\ncomputational methods to elucidate the structures and interactions\nwithin biomolecules, a task that becomes critical in the absence of\nhigh-resolution structural data. A challenging step for integrating\nthe data is knowing the expected accuracy or belief in the dataset.\nWe previously showed that the Modeling Employing Limited Data (MELD)\napproach succeeds at predicting structures and finding the best interpretation\nof the data when the initial belief is equal to or slightly lower\nthan the real value. However, the initial belief might be unknown\nto the user, as it depends on both the technique and the system of\nstudy. Here we introduce MELD-Adapt, designed to dynamically evaluate\nand infer the reliability of input data while at the same time finding\nthe best interpretation of the data and the structures compatible\nwith it. We demonstrate the utility of this method across different\nsystems, particularly emphasizing its capability to correct initial\nassumptions and identify the correct fraction of data to produce reliable\nstructural models. The approach is tested with two benchmark sets:\nthe folding of 12 proteins with coarse physical insights and the binding\nof peptides with varying affinities to the extraterminal domain using\nchemical shift perturbation data. We find that subtle differences\nin data structure (e.g., locally clustered or globally distributed),\nstarting belief, and force field preferences can have an impact on\nthe predictions, limiting the possibility of a transferable protocol\nacross all systems and data types. Nonetheless, we find a wide range\nof initial setup conditions that will lead to successful sampling\nand identification of native states, leading to a robust pipeline.\nFurthermore, disagreements about how much data is enforced and satisfied\nrapidly serve to identify incorrect setup conditions.