The Parametric Data Base

Image and molecular processing requires and produces large amounts of informational and parametric data. The traditional way to deal with this in software packages is to define one or more package-specific parameter file formats. Often, these formats have a strict notion of the placement and order of informational elements, making it hard to impossible to ensure backward compatibility as these formats evolve. The alternative to such formats is a flexible, tag-based format that is easy to maintain and extend. For Bsoft, I adopted the STAR (Self-defining Text Archiving and Retrieval) format (Hall, 1991). An alternative is XML, although as text files, XML files are less readable and takes much more space compared to the equivalent STAR files.

Internal hierarchies of structures

Bsoft was and is developed within the context of Structural Biology, and in particular associated with the use of high-resolution electron microscopy. The types of data can therefore be grouped into the image processing side, including sets of micrographs taken on electron microscopes and 3D reconstructions derived from them, and the modeling side, including atoms, molecules and molecule groups that can be arranged in lower resolution component-based models.

Micrograph parameters

The parameters derived and generated in the processing of many micrographs and their associated sub-images are encoded in an internal hierarchy as a custom database. The two main branches of the design hierachy are the field-of-view and the reconstruction. The former refers to sets of micrographs grouped in terms of their natural relationships to each other. Fields-of-view may refer to focal pairs/series, tomographic tilt series, dose series (movies), etc. The reconstruction branch is intended to reference 3D maps as from single particle or tomographic reconstructions.

Below the micrograph and reconstruction levels of the hierarchy are derived elements, such as particle, filament and marker parameters. The hierarchy preserves the relational infomartion so that the providence of derived elements are always retrievable.

Model and molecular parameters

The ultimate goal in Structural Biology is the production of a spatial model as an interpretation of the data. Bsoft provides for a hierarchical modeling scheme, with atomic models being the highest detail models, followed by succesively coarser grained models.

Additional parameter files

There are several sets of general parametric data required for various operations on atomic coordinates and sequences. These files are all located in the bsoft/parameters directory.

This file contains atomic properties such as the atomic number, the atomic weights (Pure Appl. Chem., Vol. 73, No. 4, pp. 667-683, 2001.) and the electron scattering coefficients required to calculate atomic cross-sections for electron microscopy (Peng et al., Acta Cryst. (1996). A52, 257-276). Any program reading atomic coordinates requires this file to retrieve at least the atomic weights. The program "bsf" uses the scattering curves to calculate the Fourier space representation of atomic structures.

This file contains protein residue properties such as mass, volume, charge, and hydrophobicity. In addition, it has a residue similarity matrix based on the BLOSUM62 matrix. Most of the sequence analyses requires information from this file.

This file encodes all the symmetry information for the crystallographic space groups. This was derived from the symop.lib file from the CCP4 software suite.

.krn files

These files specify kernels for convolution filters used by the program bfilter. The first line of the file gives the dimensions of the kernel. The subsequent lines give all the values of the kernel in order, with the arrangements per line being arbitrary. Here is a typical kernel for a Sobel filter in the y direction, called SobelY.krn:

3 3 3

1 3 1

0 0 0

-1 -3 -1

3 6 3

0 0 0

-3 -6 -3

1 3 1

0 0 0

-1 -3 -1