Protein Interaction Property Similarity Analysis

Frequently asked questions for PIPSA
(Protein Interaction Property Similarity Analysis)

1. I have pdb files of my proteins, but PIPSA fails to proceed them, what is wrong ?

Pdb files should prepared before running PIPSA. Depending which interaction field will be compared, prepared versions of pdb files should be readable by APBS, UHBD or GRID. Instructions for APBS and UHBD are given in usage example page example.html#pdb . Changes needed for GRID are described in GRID manual .

2. PIPSA did not superpose my proteins

3. How do I change ionic strength value to 150 mM from default 50 mM ?

Read the user guide calculation parameter section usrguide.html#param . You should modify APBS (or UHBD) preparation script: make a local copy from distribution version scr/do_pipsa_APBS_prep (or scr/do_pipsa_UHBD_pre, modify, and run (this can be done from any directory because directory locations are given as input parameters for PIPSA scripts)
Alternative is to modify apbs.in (or uhbd.in) files after the preparation script was already executed and input file for APBS (or UHBD) are created in apbs (or uhbd) work directory. This input file, apbs.in (or uhbd.in), should be edited changing ionic strength assignment. This way is useful in changing other interaction field calculation parameters locally (not changing scripts and template files in PIPSA distribution directory).

4. I have more than 999 proteins and PIPSA does not work

Read the user guide, calculation parameter section usrguide.html#param . In this case you have to change nprmx from default 999 to a larger value and recompile PIPSA programs (or at least the following programs: mkapbsin, mkdismx, mkkin, mkuhbdin, n1potsim, npotsim, qdipsim).
Changing namx to a larger value and recompiling ccenter is needed if your proteins have more than 9999 atoms.

5. I have many proteins and similarity calculations take ages

Use parallel version of PIPSA
Try to allocate grid and pdb files on the disk with faster access - PIPSA needs comparable times both for calculations and reading in data files - this is increasingly important for larger grid files.

6. I want to change one protein pdb file, should I re-calculate all pairwise similarities now ?

It is possible first remove the protein to be modified from the set of analysed proteins, using the script scr/do_pipsa_rem1 and add it again using the script scr/do_pipsa_add1. First step will only remove the protein name from the list of proteins and the entries related to it in similarity matrix file (sims.log), so that changing of pdb and grid files can be made before or after this step. The second step will use new (if updated) pdb and grid files and do calculations only for pairs involving the added protein.
Note that this method may give the analysis results different from the case when the analysis is done from the very beginning. First, the name of the replaced protein will disappear from its original location and appear at the end of the list of proteins. Second, old center for grids will be used for calculations, while this center will be recalculated and may differ from original center if all calculations are repeated.

7. How do I run PIPSA on WindowsXP / MacOSX ?

Running PIPSA on OS-s rather than Linux is not supported, but it is possible. You have to be able (a) to compile fortran programs and (b) to run pipsa shell scripts. (a) can be done using GNU compilers , a special attention to be paid -static option of compilation (at least for npotsim.f, n1potsim.f, m1potsim.f). (b) is straightforward under MacOS, can be universally solved by cygwin inder Windows, or by modifying pipsa scripts to run under respective OS.

8. Why there are 2 versions mkapbsin.f and mkapbsin-apbs-ot-0.3.2.f ?

Second program supposed to be used (instead of mkapbsin) with APBS versions older than 0.3.2. This is caused by the difference in defining UHBD format output grid origin in versions before and after 0.3.2.

9. I have electrostatic potentials as DELPHI grid files, can I compare them with PIPSA ?

You have to convert them to UHBD grid format.
The same applies to any interaction fields calculated with other programs (i.e. conversion to UHBD grid format to be done before using PIPSA).

10. How do I compare one part of my protein with its another part ?

You have to prepare 2 pdb files, the first for one part and the second for another part, superpose them so that the parts to be compared are located on the same place in 3D, and run PIPSA on these 2 proteins. For example, if you want to compare 2 monomers of a dimeric protein, the monomers should be extracted to 2 pdb files, these 2 pdb files should be superposed based on the sequence alignment, and PIPSA should be run using these 2 superposed pdb files.

11. Can I use PIPSA for non-protein molecules ?

Yes, you can, as soon as these molecules are superposed and their interaction fields are calculated and stored in UHBD grid format.

12. I can not get any correlation between the interaction fields and experimental kinetic parameters in qPIPSA analysis, what is wrong ?

According to qPIPSA methodology (here), the absence of correlation does not necessarily means that something is going wrong. Presence of correlations needs many assumptions to hold, such as:

Kinetic parameter measurement conditions should be consistent with the conditions of structural modeling. Conditions known to influence correlations are ionic strength and pH of the solvent. Either experiments under the same environmental conditions should be chosen for qPIPSA, or modeling should be adapted to each specific measurement conditions.
The proteins structures used in the analysis should be relevant for the rate-determining catalytic step. Proteins may adopt multiple conformations. Sometimes it may be necessary to do calculations with more than one conformation, e.g. with open and closed forms of an enzyme active site. If one of these gives MIFs that correlate better with known kinetic parameters, this provides somemechanistic information about the determinants of the kinetic parameter.
The region, over which interaction fields are compared, should be relevant for the kinetics. Calculations can be done for a number of regions. For TPIs, the best correlation with kinetic parameters was obtained for different regions for kcat/Km values and Km values. This again provides some mechanistic information on the parameter determinants.
The method is most suitable when the rate-determining step is mechanistically the same across the set of protein structures compared. An outlier might be mechanistically different, but if there is wide mechanistic variation in the dataset, the comparative approach cannot be expected to work.
Molecular dynamics are not currently considered in the approach and if they alter the protein structures in different ways across the dataset they will adversely affect the value of the interaction field comparison.

13. What are the two ways of running PIPSA ?

The first is to use prepared scripts (from subdirectory scr/ of pipsa distribution) combining them to accomplish the task you need. These scripts are designed to perform specific tasks taking as a parameter the data and executable locations. You may need to have local modified copies of these scripts to suit your tasks.
The second is to write scripts invoking basic programs, like n*potsim, 2potsim from pipsa's bin/ subdirectory.

Razif Gabdoulline, 2007