Deep Learning Group - A Robust Hierarchical Linear Model for Cryo-EM Map Analysis

Abstract

Cryo-electron microscopy (cryo-EM) has become a central tool for determining atomic structures of biological macromolecules, producing three-dimensional reconstruction maps that guide atomic model building. However, current quantitative analyses of cryo-EM maps typically rely on simplified Gaussian signal models with fixed width parameters, which may limit their ability to capture atom-specific signal characteristics directly from the maps. In this talk, I will introduce a robust hierarchical linear (RHL) model for the statistical analysis of paired cryo-EM maps and atomic structures deposited in the EMDB and PDB. In this framework, logarithms of local voxel intensities surrounding each atom are modeled using a linearized Gaussian form, allowing atom-specific amplitude and width parameters to be estimated through a hierarchical structure that pools information across atoms of the same type. The goal is to provide a statistically robust framework for extracting quantitative atomic signal features from cryo-EM maps, rather than imposing fixed physical assumptions on these parameters. To address contamination arising from overlapping atomic signals, spatially varying resolution, and experimental noise, we incorporate a data-adaptive weighting scheme based on minimum density power divergence estimation (MDPDE), embedded directly into the covariance structure of the model. Simulation studies show that the proposed approach produces stable parameter estimates under substantial contamination. Applications to cryo-EM maps at multiple resolutions will be presented to illustrate how the framework can reveal quantitative atomic signal patterns in high-resolution cryo-EM data.

線上視訊請點選連結