A program for fitting equilibrium binding data for homodimerization around homo-bifunctional ligands.
The mathematical foundation for this project is laid out in the publication by Mack et al.:
Mack ET, Perez-Castillejos R, Suo Z, Whitesides GM. Exact Analysis of Ligand-Induced Dimerization of Monomeric Receptors. Analytical chemistry. 2008;80(14):5550-5555. doi:10.1021/ac800578w.
This code originally written by Paul Barton while a PhD. candidate in Biological Sciences at Carnegie Mellon University.
The seminal paper for fluorogen-activating single chain antibody fragments as fluorogen-activation proteins: doi:10.1038/nbt1368 Studying the kinetics of binding for these systems provided the original motivation for creating this project.
This binding model possesses the distinct advantage of having an analytical solution, providing the ability to exactly (without simplifying approximations) compute how the system should behave theoretically under varied conditions. However the computation of this model faces a particular challenge (one readily demonstrated in the notebook of this project) in that it is highly sensitive to floating point imprecision depending on the parameters of the system/experiment. Put another way, with the right circumstances, a computation of the model may yield a wildly inaccurate answer or fail utterly.
The response to this challenge involves utilizing ever more precise representations of floating point numbers such that the imprecision becomes irrelevant to the system, and its parameters, at hand. With this tactic in mind I began to write the code to fit experimental data to this binding model using SciPy and mpmath. The first providing an tools for solving error minimization problems in curve fitting, and the latter allowing arbitrary levels of precision in floating point numerical operations. This ultimately failed because, as I came to learn, SciPy was only providing a wrapper around various mathematical packages, often compiled fortran, which were not readily able to interact with mpmath.
It then appeared that a reasonable strategy would be to adapt the fortran packages I would need to employ for direct use with arbitrary floating point precision. My research into this problem led me to employ MPFUN2015 by David H. Bailey (http://www.davidhbailey.com/dhbsoftware/) as I adapted the necessary subroutines from MINPACK so that they might successfully fit experimental data with the model.
Fortran lies at the heart of this project and I am using Fortran90/95. MPFUN2015 is a Fortran90 package, and the MINPACK from which I worked is also Fortran90. To successfully use this project, you will need an adequate Fortran compiler.
Python plays an important role in this project, I am using Python3 and a myriad of packages not in the standard library for various jobs. To use the core function of ligfit (fitting data!), you will need NumPy and docopt. I won't remark on the dependencies of NumPy here, but I will note that as of the time of writing (2015-05-24) I had to install the traditional way rather than using pip, as installing via pip did not also install the vital tool f2py. The traditional way being: first download the source, then compile/install with "python(3) setup.py install".
To make use of the notebook and the other various adjuncts I have provided, you will need to install ipython, scipy, and matplotlib. Pip for these should do, the setup.py installation script for this project should automatically get these dependencies for you.