LovoFit

Institute of Chemistry - University of Campinas

Home Usage
Performs a linear fitting of (perhaps multidimensional) data to target values, automatically eliminating outliers. Needs lapack.

Compile with: gfortran -o lovofit lovofit.f90 -llapack
Run with: lovofit ./data.dat > fit.dat

where data.dat contains the data file in the form:
          2 135 0.8
          [Y1] [x11] .... [x1N]
          [Y2] [x11] .... [x2N]
          ...
where '2' is the number of parameters (the dimension of the domain plus one), '135' is the number of data points, and '0.8' is the fraction of the data to be explicitly considered in the alignment (in this case 20% of the data will be considered as outliers). [Y1] are the target values (data to be fitted) and [X..] are the domain points. The output file will contain, in the header, the parameters of the fit, and a table of the predicted data vs. the given data. The final column will be an indicator of the usage, or not, of that point to the parameter estimation. Run with the -print 2 option to automatically generate a xmgrace plot.
See also:
Packmol
LovoAlign
MDLovoFit
Author's software page
The TANGO project
 

Tutorial

1. Description of MDLovoFit capabilites and expected results.
    Read this to understand what to get from the package.
2. Step by step tutorial on how to use the package in your simulation.
    Follow the tutorial and get the results from your simulation.


1. Description of MDLovoFit capabilites and expected results.


First, we will show which kind of information it is expected from the usage of this package. Two plots are characteristic of the analysis of the data performed by MDLovoFit. Figure 1 illustrates the analysis of the mobility of a protein structure in a typical MD simulation.


Figure 1

The panel A of Figure 1 shows the standard RMSD of the protein Cα atoms along the simulation. It indicates that the protein somewhat diverges from the initial structure.

Figure 1B is the result of running MDLovoFit with the -mapfrac option. It shows the RMSD of the RMSD of the alignment of a fraction of the atoms, as a function of this fraction. It shows that, for example, it is possible in this simulation to align about 80% of the Cα atoms of the structure to less than 1Å

Finally, in figure 1C, the RMSD computed by MDLovoFit for the alignment of 70% of the Cα atoms is shown. Clearly, the fluctuations of 70% of the structure are minimal, while there is a small subset of atoms which deviates from the initial structure and explains, by themselves, the divergence observed in figure 1A.

The result of figure 1C can, of course, be supported by the visualization of the resulting alignment. Figure 2, which is also obtained from the output of MDLovoFit (and using VMD)


Figure 2

This figure displays visually, in blue, the 70% of the Cα atoms of the protein which were automatically identified by MDLovoFit as the least mobile ones (with RMSD lower than 1Å, as shown in figure 1C). In red, the figure highlights which regions of the structure are more mobile, and diverge from the initial conformation of the simulation.

With the plots of figures 1B and 1C, and with the above structural representation, a more comprehensive view of the mobility of the protein can be obtained. In the next section we will show how to, in practice, obtain each of this plots.

Additionally, it is possible to plot the RMSF of the atoms, and to use this data to color the structure, using VMD. See the last section of the tutorial (next section) for details.
See also:
Packmol
LovoAlign
MDLovoFit
Author's software page
The TANGO project