MaD: Macromolecular Descriptors for Integrative Dynamic Modeling Supported by Cryo-EM Data

Abstract

The determination of atomic structures of large flexible systems remains a challenging task despite the recent advances in cryo-electron microscopy (cryo-EM) and de novo protein structure prediction. Few hybrid methods truly consider dynamics because deriving appropriate conformational ensembles and scoring functions robust to conformational differences is not trivial. We present here Macromolecular Descriptors (MaD) which, inspired by traditional computer vision concepts, remedies some of these limitations. MaD identifies feature points and encodes local structural information around them into resolution- and conformation-invariant descriptors. Efficient matching of descriptors from cryo-EM densities at low and medium resolution with those of high-resolution component structures yields a robust and accurate assembly prediction that does not require other experimental or computational input. Fast, scalable and easy to use, this method is able to incorporate native dynamic features as extracted from molecular simulations and identify the models that best match the target electron density. Therefore, MaD addresses some of the unanswered needs of the community in terms of the integrative modeling of large and flexible macromolecular complexes.