Last Modified: Thursday, 23-Oct-2008 15:51:52 EDT

2008.10.23 -- Questions about the Validity of OS output

It is observed that using the test data in the OS distribution, different results are obtained with V6 (SGI) and V7.2.3 (PC/linux). (see Bottom Line.)

The differences occur in the first step, generating the dot surface. The program ds/ds72 is repeatedly invoked from main/main72, which is the first program invoked by the os.run script.

ds reads part_i.pdb and generates part_i.ms.

part_i.pdb is a fractional pdb file, containing all the atoms in a single residue, plus the N from the next residue, and the C from the previous residue.

part_i.ms describes the molecular surface of these atoms. See examples below. For each atom, there is an atom record, followed by records for each dot on the atom's surface. Area is the dot's area. Code, for linux, is ST followed by a single digit (dot_ISHAPE) which is probably a code for convex, concave, saddle, etc.

                |- atom/dot coords ---| code area  |-surface normal --| 
First 5 lines SGI:
ARG    1  CA    5.320    5.998  -10.170 A
ARG    1  CA    3.449    6.234   -9.941 SC0  0.214 -0.985  0.124  0.121
ARG    1  CA    3.543    5.998   -9.496 SC0  0.214 -0.935  0.000  0.355
ARG    1  CA    3.604    5.538   -9.496 SC0  0.214 -0.903 -0.242  0.355
ARG    1  CA    3.604    6.458   -9.496 SC0  0.214 -0.903  0.242  0.355

First 5 lines Linux:
ARG    1  CA    5.320    5.998  -10.170 A
ARG    1  CA    3.402    6.070  -10.556 ST3  0.186 -0.957  0.278  0.087
ARG    1  CA    3.472    6.446  -10.704 ST3  0.184 -0.982  0.129  0.140
ARG    1  CA    3.433    6.206  -10.592 ST2  0.056 -0.967  0.241  0.087
ARG    1  CA    5.196    4.113   -9.744 ST3  0.101 -0.002 -0.871  0.491

1. ds72 has 9620 lines of code vs 3128 for ds
2. ds appears to assign equal size to all dots on a given atom. ds72 calculates an individual size for each dot.
3. The two versions do not produce the same number of dots for each atom.
4. V6 appears to generate equi-space dots by moving in circles at fixed z-intervals. V7.3.2's algorithm cannot be easily deduced by looking at the output.
5. They use different codes to indicate the type of each dot (convex, concave, saddle, etc.).

It is hardly surprising that they give different results. One suspects that 7.2.3 gives better results because of a more sophisticated treatment of each surface dot.

The results of subsequent steps in the OS script also differ between the two versions. This is hardly surprising, because they are based on the .ms filers generated by ds/ds72.

Bottom Line

Back to OS Description

Last Modified: Thursday, 23-Oct-2008 15:51:52 EDT