Brief

Researchers have developed CryoProt, a new framework for pretraining protein representations using cryo-electron microscopy (cryo-EM) density maps. This method addresses the limitation of existing approaches by explicitly modeling interactions between different regions of the density map, rather than treating them independently. CryoProt utilizes a multi-task pretraining strategy to learn generalizable representations that can be applied to various downstream tasks, showing significant improvements in protein flexibility prediction. AI

Researchers have developed a novel data fusion framework called the method of double moments (MoDM) for reconstructing 3-D molecular structures from cryo-electron microscopy images. This method leverages two distinct sets of second-order moment statistics derived from projection images captured under different orientation distributions. The framework proves that these moments can uniquely determine the molecular structure, and an algorithm based on convex relaxation achieves accurate recovery using only these second-order statistics. AI