ibp-ng

An implementation (and experimentation sandbox) of the interval
Branch-and-Prune (iBP) structure determination algorithm.

Introduction

The iBP algorithm enumerates the set of molecular structures that conform
to a user-specified set of distance, angle and dihedral constraints. For
a proof of principle in biomolecules, the reader is referred to:

Cassioli, et al., An algorithm to enumerate all possible protein
conformations verifying a set of distance constraints,
BMC Bioinformatics, 2015, 16: 23.

The mathematics used to embed each atom into three-dimensional space are
discussed in:

Lavor et al., Clifford algebra and the discretizable molecular distance
geometry problem, Advances in Applied Clifford Algebras,
2015, 25: 925.

A recursive description of the iBP algorithm is published in:

Lavor et al., The interval Branch-and-Prune algorithm for the
discretizable molecular distance geometry problem with inexact
distances, Journal of Global Optimization, 2013, 56: 855.

The current implementation uses both breadth-first and depth-first iterative
tree searches based on a multidimensional index data structure, but this may
change in future versions.

Compilation

The iBP source code should be compilable on any decently modern
GNU/Linux distribution. The development configuration is:

Linux 2.6.18
Binutils 2.17.50
GCC 4.1.2
Flex 2.5.4
Bison 2.3

There are two compilation options that must be set in the
[Makefile] (Makefile):

IBP_PTHREAD: enable support for multiple parallel (CPU) threads.
IBP_CUDA: enable (experimental, incomplete) support for GPU threads.

Once these options are set, ibp-ng may be compiled like so:

make

A simple pull-and-build for the lazy:

git clone git://github.com/geekysuavo/ibp-ng.git
cd ibp-ng
mkdir bin
make

Basic Usage

Successful compilation will produce a file in bin called ibp-ng.
For basic usage information, just run:

bin/ibp-ng -h

A set of examples is provided in data. Within each example in
data, the script that invokes ibp-ng is named run, e.g.
data/alpha/run. In addition, a set of test cases is
available in the expts directory, where one can run run.py
in order to prepare the input files required to run ibp-ng:

cd expts
python run.py

More examples will be placed in the data directory as the source
code progresses.

Licensing

This project is released under the
MIT license. See the
LICENSE.md file for the complete license terms.