(this code is under active development, and nothing works... yet!)
Structural ensemble determination from heterogeneous, multiple experimental sources.
ODIN takes your experimental data and predicts structures, and also structural ensembles in a rigorous way!
ODIN is a flexible software package that implements statistical mechanical theory capable of infering geometrical structures from almost any experiment. The main focus of the software is on biomolecules studied via x-ray scattering and NMR, but in principle could be adapted to a number of different experimental systems and techniques.
The driving idea behind ODIN is that when an experiment is performed on a system, that experiment contains some quantity of information about the system's geometry. Extracting that information may be difficult. If, however, we possess an algorithm for predicting experiments from structures (often a much easier problem), then a direct comparison is possible. ODIN employs this comparison to generate a structural ensemble that, when run through such an algorithm, is guarenteed to reproduce the experimental data. Further, ODIN works to ensure that this ensemble contains as little "information", as measured by the entropy of the ensemble's Boltzmann distribution, as possible.
Details of the theory and ideas behind the inner workings of ODIN can be found in the references, below, and documentation included in the package.
In addition to providing structure determination functionality, ODIN contains:
- Powerful tools for predicting SAXS and WAXS experiments, and analyzing these experiments using traditional and correlation techniques.
- python2.7+ (epd)
- numpy (epd)
- scipy (epd)
- pytables (epd)
- pyyaml (epd)
- matplotlib (epd)
- OpenMM (https://simtk.org/home/openmm)
- periodictable (http://www.reflectometry.org/danse/elements.html)
- nvcc (optional additional GPU support, https://developer.nvidia.com/cuda-downloads)
We HIGHLY recommend using the Enthought python distribution (EPD, http://www.enthought.com/products/epd.php) which contains all of the dependincies marked above with (epd) in a single package. It's awesome and free for academics.
We support Linux exclusively. MacOS and Windows should work, they are just largely untested. If someone wants to try on either, or has success, please let us know: tjlane@stanford.edu.
- Install the above dependencies.
- Run
python setup.py install - Test
python setup.py test(optional) - Enjoy!
- TJ Lane
- Derek Mendez
- Yutong Zhao
- Robert McGibbon
ODIN is full-GPL. See LICENCE file.
Various parts of ODIN are derived from, or explicitly borrow from:
- Robert McGibbon & Yutong Zhao's sweet GPU-to-python wrapping scheme: https://github.com/rmcgibbo/npcuda-example
- MSMBuilder & OpenMM PDB & trajectory classes: https://github.com/rmcgibbo/mdtraj
- grid: https://github.com/dermen/grid