A convolutional autoencoder approach for mining features in cellular electron cryo-tomograms and weakly supervised coarse segmentation

Abstract

Cellular electron cryo-tomography enables the 3D visualization of cellular organization in the near-native state and at submolecular resolution. However, the contents of cellular tomograms are often complex, making it difficult to automatically isolate different in situ cellular components. In this paper, we propose a convolutional autoencoder-based unsupervised approach to provide a coarse grouping of 3D small subvolumes extracted from tomograms. We demonstrate that the autoencoder can be used for efficient and coarse characterization of features of macromolecular complexes and surfaces, such as membranes. In addition, the autoencoder can be used to detect non-cellular features related to sample preparation and data collection, such as carbon edges from the grid and tomogram boundaries. The autoencoder is also able to detect patterns that may indicate spatial interactions between cellular components. Furthermore, we demonstrate that our autoencoder can be used for weakly supervised semantic segmentation of cellular components, requiring a very small amount of manual annotation.

An integration of fast alignment and maximum-likelihood methods for electron subtomogram averaging and classification

“…The extracted subtomograms are highly heterogeneous. Therefore, we used a convolutional autoencoder ( Zeng et al , 2017 ) to perform unsupervised clustering of the extracted subtomograms and selected only the clusters with large globular features because they are more likely to be ribosomes. This filtering process selected about 10% subtomograms for further analysis.…”

“…The limitations of the averaging methods are also carried to the extended classification tasks. To reduce the heterogeneity of millions of structurally highly diverse macromolecules, we have developed deep learning based unsupervised classification method ( Zeng et al , 2017 ) that can coarsely group subtomograms into more homogeneous clusters without accurate alignment. Clusters of interest can be selected for further analysis.…”

Template-Based and Template-Free Approaches in Cellular Cryo-Electron Tomography Structural Pattern Mining

“…A convolutional autoencoder-based unsupervised approach has been proposed (25) for coarse selection of subtomograms of interest from a large number of subtomograms (scale ranging from thousands to millions). After subtomograms are extracted from the tomogram using particle picking methods, an optional pose normalization approach is provided to adjust the particle orientation and center for better clustering of the same structures of different orientations, which simplifies the process of structural mining.…”

Tomographic image correction with noise reduction algorithms

“…The assumption of using this kind of denoise function is that good representation is immune to interference and interference of input data. Figure 2 presents the model of the convolutional autoencoder stacked together with multilayer perceptron (MLP) in the last layer of the stack [20]. The stack consists of three encoders and one regressive artificial neural network MLP (multilayer perceptron) type.…”