3D Structure from 2D Microscopy images using Deep Learning

TL;DR

This paper introduces a deep learning approach to reconstruct 3D protein structures from 2D microscopy images, overcoming limitations of traditional voxel-based methods and providing accurate models for complex biological structures.

Contribution

The authors develop a novel convolutional neural network with a differentiable renderer that predicts protein complex structures directly from 2D microscopy images, without constraints.

Findings

Successfully reconstructed CEP152 protein complex structure.

Achieved accurate 3D models of centrioles.

Demonstrated effectiveness on two biological datasets.

Abstract

Understanding the structure of a protein complex is crucial indetermining its function. However, retrieving accurate 3D structures from microscopy images is highly challenging, particularly as many imaging modalities are two-dimensional. Recent advances in Artificial Intelligence have been applied to this problem, primarily using voxel based approaches to analyse sets of electron microscopy images. Herewe present a deep learning solution for reconstructing the protein com-plexes from a number of 2D single molecule localization microscopy images, with the solution being completely unconstrained. Our convolutional neural network coupled with a differentiable renderer predicts pose and derives a single structure. After training, the network is dis-carded, with the output of this method being a structural model which fits the data-set. We demonstrate the performance of our system on two protein complexes: CEP152 (which comprises part of the proximal toroid of the centriole) and centrioles.