Deep Learning Improves Macromolecule Identification in 3D Cellular Cryo-Electron Tomograms

Abstract Cryo-electron tomography (cryo-ET) visualizes the 3D spatial distribution of macromolecules at nanometer resolution inside native cells. While this label-free cryogenic imaging technology produces data containing rich structural information, automated identification of macromolecules inside cellular tomograms is challenged by noise and reconstruction artifacts, as well as the presence of many molecular species in the crowded volumes. Here, we present a computational procedure that uses artificial neural networks to simultaneously localize with a multi-class strategy several macromolecular species in cellular cryo-electron tomograms. Once trained, the inference stage of DeepFinder is significantly faster than template matching, and performs better than other competitive deep learning methods at identifying macromolecules of various sizes in both synthetic and experimental datasets. On cellular cryo-ET data, DeepFinder localized membrane-bound and cytosolic ribosomes (~3.2 MDa), Rubisco (~540 kDa soluble complex), and photosystem II (~550 kDa membrane complex) with comparable accuracy to expert-supervised ground truth annotations. Furthermore, we show that DeepFinder is flexible and can be combined with template matching to localize the missing macromolecules not found by one or the other method. The DeepFinder algorithm is therefore very promising for the semi-automated analysis of a wide range of molecular targets in cellular tomograms, including macromolecules with weights of 500-600 kDa and membrane proteins.