---------------------------------------------------------------------- The ROME (Refinement and Optimization via Machine lEarning for cryo-EM) software package is a parallel computing software system dedicated for high-resolution cryo-EM structure determination and data analysis, which implements advanced machine learning approaches in modern computer sciences and runs natively in an HPC environment. The ROME 1.0 introduces SML (statistical manifold learning)-based deep classification following MAP-based image alignment. It also implemented traditional unsupervised MAP-based classification and includes several useful tools, such as 2D class averaging with CTF (contrast transfer function) correction and a convenient GUI for curation, inspection and verification of single-particle classes. The ROME system has be optimized on both Intel® Xeon multi-core CPUs and Intel® Xeon Phi many-core coprocessors. For more information, please visit: https://github.com/AlphaCryo4D/rome ---------------------------------------------------------------------- ROME is an OPEN SOURCE program distributed under the General Public license. Please see the LICENSE file for additional information. ---------------------------------------------------------------------- See INSTALL for installation instructions ---------------------------------------------------------------------- ROME has the following external dependencies: 1. Intel MKL ROME GUI has the following external dependencies: 1. Python 2. Pillow 3. numpy 4. scipy ---------------------------------------------------------------------- The ROME Team: NOTE: though IPCCSB specifically develops and maintains the ROME code base, the project can only continue to move forward through the sponsorship and participation of the broader community. Founder and Principal Author: Youdong Jack Mao Algorithmic Implementation: Jiayi Wu Yong-Bei Ma Yingping Ma Code Optimization: Charles Condgon Bevin Brett ---------------------------------------------------------------------- Disclaimer: The source codes of ROME either contain or are derived from pieces of codes from the following packages. RELION: http://www2.mrc-lmb.cam.ac.uk/relion/ XMIPP: http:/xmipp.cnb.csic.es BSOFT: http://lsbr.niams.nih.gov/bsoft/ SPIDER: http://spider.wadsworth.org/ HEALPIX: http://healpix.jpl.nasa.gov/ Original disclaimers in the code of these external packages have been maintained as much as possible. Please contact Youdong Mao (Youdong_Mao@DFCI.HARVARD.EDU) if you feel this has not been done correctly. ---------------------------------------------------------------------- ROME 1.1.0 (March 29, 2017): Unsupervised deep 2D classification is introduced. It implementes: 1) Adaptive MAP-based image alignment and classification; 2) SML based on GTM adapted to Fourier space; 3) Deep2D protocol that combines MAP alignment with SML classification. The SML/Deep2D principle and method are documented in: J. Wu, Y. Ma, C. Condgon, B. Brett, S. Chen, Q. Ouyang, Y. Mao. Unsupervised single-particle deep classification via statistical manifold learning. arXiv:1604.04539 [physics.data-an] (2016). Unsupervised 3D classification based on adaptive MAP method is implemented with optimization toward many-core CPU architecture. The adaptive MAP method was documented in: Sigworth, F.J. A maximum-likelihood approach to single-particle image refinement. J Struct Biol 122, 328-339 (1998). Scheres, S.H. Maximum-likelihood Multi-reference Refinement for Electron Microscopy Images. J Mol Biol 348, 139-149 (2005). Tagare, H.D., Barthel, A. & Sigworth, F.J. An adaptive Expectation-Maximization algorithm with GPU implementation for electron cryomicroscopy. J Struct Biol 171, 256-265 (2010). Scheres, S.H. A Bayesian view on cryo-EM structure determination. J Mol Biol 415, 406-418 (2012). ROME 1.1.2 (March 29, 2019): This minor update adds a few new programs based on previous version: 1) A C++ implementation of ResMap with parallelization; 2) A parallelized implementation of standalone 3D reconstruction program; The following bugs have been addressed at least partly in this new version: 1. Minor memory issue during loading a large input stack file is solved; 2. Fixed some input and output bugs; 3. Fixed some OMP and memory bugs; 4. Fixed an image read bug and a metadata access bug; 5. Select direction from prior angle; 6. Fixed some bugs in the reconstruction program; 7. Added Mask option in ROME 2D and 3D classification program; 8. Bypassed some local searching and only select non-zero probabilities while sorting, to speedup the 3D reconstructions; 9. Removed ScaleCorrection output in star file, and added all relion metadata labels, to be compatible with RELION. ---------------------------------------------------------------------- Specific Acknowledgments: * Thanks to Kristina M. Kermanshache for advocate of Intel sponsorship, collaboration, corporate coordination and many help in project development. * Thanks to Mark Begley and Michael Moretti for help in obtaining Intel software development tools. * Thanks to Ram Ramanujam for project coordination. * Thanks to Clay Breshears and David S. Scott for helpful discussion. * Thanks to Hao Wu, Liman Zhang and Yang Li in early adoption of ROME in their research work. * Thanks to Shuobing Chen and Yanan Zhu for help in testing ROME. * Thanks to Jonathon Jackson and Tao Song in configuration and maintenance of the testing cluster system.
AlphaCryo4D/ROME
The ROME (Refinement and Optimization via Machine lEarning for cryo-EM) software package is a parallel computing software system dedicated for high-resolution cryo-EM structure determination and data analysis, which implements advanced machine learning approaches in modern computer sciences and runs natively in an HPC environment. The ROME 1.0 introduces SML (statistical manifold learning)-based deep classification following MAP-based image alignment. It also implemented traditional unsupervised MAP-based classification and includes several useful tools, such as 2D class averaging with CTF (contrast transfer function) correction and a convenient GUI for curation, inspection and verification of single-particle classes. The ROME system has be optimized on both Intel® Xeon multi-core CPUs and Intel® Xeon Phi many-core coprocessors.
C++GPL-3.0