Enabling Deep Geometric Learning on Cryo-EM Maps Using Neural Representation

Abstract

Advances in imagery at atomic and near-atomic resolution, such as cryogenic electron microscopy (cryo-EM), have led to an influx of high resolution images of proteins and other macromolecular structures to data banks worldwide. Deep geometric learning is intriguing for use in structure segmentation, but the native voxel format of cryo-EM maps is unsuitable as input to such methods. We present a novel data format called the neural cryo-EM map that accurately parameterizes cryo-EM maps and provides native, spatially continuous density and gradient data to serve as the basis for a graph-based interpretation of cryo-EM maps. Density values interpolated using the non-linear neural cryo-EM format are more accurate than conventional tri-linear interpolation. Our graph-based interpretations of 115 experimental cryo-EM maps from 1.15 to 4.0 Angstrom resolution provide high coverage of the underlying amino acid residue locations, while accuracy of nodes is correlated with resolution. The nodes of graphs created from atomic resolution maps (higher than 1.6 Angstrom) provide greater than 99% residue coverage as well as 85% full atomic coverage with a mean of 0.19 Angstrom root mean squared deviation (RMSD). Other graphs have a mean 84% residue coverage with less specificity of the nodes due to experimental noise and differences of density context at lower resolutions. Graphs created from atomic resolution maps may serve as input to downstream deep geometric learning applications and may be generalized to transform any 3D grid-based data format into non-linear, continuous, and differentiable format.