Domain Randomization for Macromolecule Structure Classification and Segmentation in Electron Cyro-tomograms

It is crucial to study and understand cellular processes. In recent years, Cellular Electron CryoTomography (CECT) serves as a powerful 3D imaging tool to visualize spatial structure of macromolecules inside the cell. However, it is challenging to analyze the macromolecular structures in a systematic way due to nature of the structural complexity of subcellular components. Existing computational and deep learning based approaches suffer from limited scalability, discrimination ability and lack of accurate annotated CECT data. Training with cheap simulated data can alleviate this problem while facing new challenges of bridging the “reality gap” between synthetic training data and real testing data. In this paper, we tackle the tasks of macromolecule structure classification and segmentation in CECT images by adapting a simple but effective technique, domain randomization. We show that by combining deep neural models and domain randomization, we are able to achieve significant improvements of 35.21% and 46.34% in tasks of classification and semantic segmentation for real CECT data, comparing to the model trained only on syhthetic data that aims to faithfully reproduce real-world data distribution.

[1]  Pieter Abbeel,et al.  Superhuman performance of surgical tasks by robots using iterative learning from human-guided demonstrations , 2010, 2010 IEEE International Conference on Robotics and Automation.

[2]  J. Frank.  Three-Dimensional Electron Microscopy of Macromolecular Assemblies , 2006 .

[3]  G Sapiro,et al.  Classification and 3D averaging with missing wedge correction in biological electron tomography. , 2008, Journal of structural biology.

[4]  Catherine M. Oikonomou,et al.  Cellular Electron Cryotomography: Toward Structural Biology In Situ. , 2017, Annual review of biochemistry.

[5]  Barbara Caputo,et al.  Learning the Roots of Visual Domain Shift , 2016, ECCV Workshops.

[6]  Luca Antiga,et al.  Automatic differentiation in PyTorch , 2017 .

[7]  R. Henderson,et al.  Detective quantum efficiency of electron area detectors in electron microscopy , 2009, Ultramicroscopy.

[8]  Yoshua Bengio,et al.  How transferable are features in deep neural networks? , 2014, NIPS.

[9]  Karim Elmaaroufi,et al.  Improved deep learning-based macromolecules structure classification from electron cryo-tomograms , 2017, Machine Vision and Applications.

[10]  Trevor Darrell,et al.  LSDA: Large Scale Detection through Adaptation , 2014, NIPS.

[11]  Xiao-Ping Xu,et al.  Efficient Extraction of Macromolecular Complexes from Electron Tomograms Based on Reduced Representation Templates , 2015, CAIP.

[12]  Eric P. Xing,et al.  Deep Learning Based Supervised Semantic Segmentation of Electron Cryo-Subtomograms , 2018, 2018 25th IEEE International Conference on Image Processing (ICIP).

[13]  Frank Alber,et al.  High-throughput subtomogram alignment and classification by Fourier space constrained fast volumetric matching. , 2012, Journal of structural biology.

[14]  M. Topf,et al.  Two distinct trimeric conformations of natively membrane-anchored full-length herpes simplex virus 1 glycoprotein B , 2016, Proceedings of the National Academy of Sciences.

[15]  J. Briggs.  Structural biology in situ--the potential of subtomogram averaging. , 2013, Current opinion in structural biology.

[16]  Pieter Abbeel,et al.  Using inaccurate models in reinforcement learning , 2006, ICML.

[17]  Ross B. Girshick,et al.  Mask R-CNN , 2017, 1703.06870.

[18]  Thomas Brox,et al.  U-Net: Convolutional Networks for Biomedical Image Segmentation , 2015, MICCAI.

[19]  J. Mccammon,et al.  Situs: A package for docking crystal structures into low-resolution maps from electron microscopy. , 1999, Journal of structural biology.

[20]  Friedrich Förster,et al.  TOM software toolbox: acquisition and analysis for electron tomography. , 2005, Journal of structural biology.

[21]  Jiaying Liu,et al.  Revisiting Batch Normalization For Practical Domain Adaptation , 2016, ICLR.

[22]  W. Baumeister,et al.  In Situ Structure of Neuronal C9orf72 Poly-GA Aggregates Reveals Proteasome Recruitment , 2018, Cell.

[23]  Eric P. Xing,et al.  Deep learning-based subdivision approach for large scale macromolecules structure recovery from electron cryo tomograms , 2017, Bioinform..

[24]  Wojciech Zaremba,et al.  Domain randomization for transferring deep neural networks from simulation to the real world , 2017, 2017 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS).

[25]  Trevor Darrell,et al.  Efficient Learning of Domain-invariant Image Representations , 2013, ICLR.

[26]  M. Valle,et al.  Averaging of electron subtomograms and random conical tilt reconstructions through likelihood optimization. , 2009, Structure.

[27]  R. Aebersold,et al.  Visual proteomics of the human pathogen Leptospira interrogans , 2009, Nature Methods.

[28]  Neil D. Lawrence,et al.  Dataset Shift in Machine Learning , 2009 .

[29]  Wolfgang Baumeister,et al.  A visual approach to proteomics , 2006, Nature Reviews Molecular Cell Biology.

[30]  Friedrich Förster,et al.  Classification of cryo-electron sub-tomograms using constrained correlation. , 2008, Journal of structural biology.

[31]  Min Xu,et al.  Automated multidimensional phenotypic profiling using large public microarray repositories , 2009, Proceedings of the National Academy of Sciences.

[32]  Trevor Darrell,et al.  What you saw is not what you get: Domain adaptation using asymmetric kernel transforms , 2011, CVPR 2011.