User guide to STAPAC and PIPSA

Often, kinetic parameters for enzymes such as K_m or k_cat/K_m are not available from the desired species but have been measured for enzymes from other species.

Sometimes the parameters for the enzyme of interest were measured under different environmental conditions such as pH, ionic strength etc. The user may be interested in characterizing possible enzyme variants such as mutants, different isoforms etc. Sometimes several mesurements of a kinetic parameter are available but these differ significantly and a parameter needs to be selected.

The electrostatic potential is one of the key determinants of enzymatic catalysis. PIPSA, Protein Interaction Property Similarity Analysis [1,2], can be used to aid the estimation of kinetic parameters from a similarity analysis of the electrostatic potentials of the enzyme for which the parameter is needed and the enzymes for which parameters are already known.

Experimentally determined kinetic parameters for the enzyme given by the EC annotation of the given Swiss-Prot entry are automatically extracted from BRENDA ( http://www.brenda-enzymes.info) and SABIO-RK (http://sabio.villa-bosch.de/SABIORK) databases by STAPAC. Links to related Swiss-Prot entries found in BRENDA and SABIO-RK are added to the PIPSA analysis.

Current limitations are:

1. Only single chain proteins can be modelled
2. Neglect of protein flexibility
3. Lack of user intervention in PIPSA

PIPSA incorporates the following steps:

1. Retrieval of related protein sequences from Swiss-Prot (http://au.expasy.org) and of structure files from PDB (http://www.rcsb.org), MODBASE (http://modbase.compbio.ucsf.edu/) and Swiss-Model (http://swissmodel.expasy.org//SWISS-MODEL.html).
2. Choice of suitable experimental structure from the PDB to use as a template for the generation of protein structural models by homology modelling. Please expand the respective section on the page by clicking on the arrow.
3. Automatic generation of protein structural models for the enzyme of interest and all those enzymes for which kinetic parameters are available using the chosen PDB template structure.
4. Calculation of the electrostatic potentials for all protein structural models.
5. Comparison of the electrostatic potentials and calculation of similarity indices for all protein structural models. The user has the option to
   1. compare the electrostatic potential around the entire protein or
   2. restrict the region of comparison to a region around the active site or other regions based on the SwissProt entry. The selection of the protein regions can be done after expanding the respective section on the page (Please click on the arrow). All features with the same Swiss-Prot key will be assemblied in one region that is treated in one comparison. If you have two distinct binding regions, it is advisable to rename the key of one region to allow separate processing of the two regions.
6. Display of the relations between the enzymes compared in electrostatic potential distance space in an ‘epogram’.

STAPAC will retrieve protein structures from PDB or MODBASE and Swiss-Model that possess the same Enzyme Classification (EC) number as the SwissProt entry you entered.

Non-experimental, i.e. theoretically modelled, protein structures from the Modbase [2] (http://modbase.compbio.ucsf.edu) or Swissmodel Repository (http://swissmodel.expasy.org/repository) [3] will be downloaded in case no PDB structure was found.

Points to consider for the selection of the template:

1. Coverage and identity level of how much of your sequence is covered by the PDB structure (excluding all gaps in the alignment) as well as level of amino acid sequence identity between Swiss-Prot entry and PDB sequences. These numbers are given in the first column as a maximum of all chains in an entry.
2. Chains may differ in coverage and identity. Also one needs to make sure that important features of your sequence are covered by the PDB structure. The chains column gives information about the origin of the sequence and the residues involved. Also the coverage, identity and t-coffee (http://www.tcoffee.org) alignment score (higher score indicate better alignments) is given.
3. RMSD (root mean square deviation) of each template to the best template gives a hint of how different the templates are. A low RMSD value indicates that the two structures are almost identical. STAMP is used for the multiple structural alignment. The domains of each Struture are extracted and the aligned all-against-all. You can see the STAMP logfile by clicking at the RMS value.
4. Details of the structure to consider:
   • Resolution of X-ray structure and completeness of structure.
   • Conformation and state of the crystallized enzyme (the user may have to go to the primary literature).
   • Whether the reaction mechanism is identical (if known, the user may have to go to the primary literature)
   • Experimental conditions under which the structure was solved, e.g. pH, buffer etc.
   • Number of missing residues in the pprotein structure. To see the whole list of missing residues click on the triangle.

MODELLER [1] will be used to generate 3D protein structural models for

1. the SwissProt entry for which you want to estimate kinetic parameters
2. the enzymes for which kinetic parameters were found in BRENDA or SABIO-RK
3. the amino acid sequence of the template (for control and consistency checks).

A MODELLER protocol is applied that ensures the reproduction of the side-chain orientation of conserved amino acid residues. It consists of the following -steps:

1. A ClustalW [2] multiple sequence alignment will be performed of the protein amino acid sequences to the PDB entry (used as a template structure in the following homology modeling).
2. The amino acid sequence of the SwissProt entry will be adjusted in length to match the sequence covered by the PDB entry.
3. Generation of initial models based on structural constraints from the PDB template structure.
4. Energy minimization of the generated homology models.

The program UHBD [1] is used to calculate the electrostatic potentials for all protein structural models. The default settings include

1. Maximum grid size dimension 110x110x110 grid points (an automatic screening of grid size dimensioning to ensure protein coverage is performed).
2. Spacing between grid points 1.0 Å.
3. Ionic strength 50 mM.
4. Protein dielectric constant: 4, solvent dielectric constant: 78.
5. The enzyme molecular surface is defined as the molecular surface being accessible to a spherical water probe of 1.4 Å.
6. Skin thickness of 3 Å and a probe radius of 2 Å (for comparing the electrostatic potentials in a region of particular interest, such as active site).

PIPSA [1,2] will be used to perform a pairwise comparison of the electrostatic potentials of the protein structural models. The user has to decide whether to perform the analysis for

1. The entire protein (this will be performed in all cases).
2. Restrict the comparative analysis to a region (sphere) around residue(s) that are annotated in the Swiss-Prot database as
   1. ACT_SITE
   2. BINDING
   3. etc. all Swiss-Prot features can be selected

The pairwise similarities of the electrostatic potentials of the SwissProt entry, the enzymes with kinetic data available from BRENDA and SABIO-RK and the PDB template structure are displayed as a phylogenetic tree-like diagram (“epogram”).

The user may choose the kinetic parameter of the closest related enzyme. The reliability of this estimate depends on a number of factors such as conformational flexibility of the enzyme, appropriateness of the electrostatic potential to describe the enzyme kinetics etc. and use it with care in the mathematical modeling of biochemical network.

This result of a qPIPSA a tree-like diagram ("dendogram") which puts the electrostatic potential of the query sequence into relation with other enzymes for which kinetic data are available. In PIPSA, the distance matrix (SQRT(1-SI)) of the electrostatic potential is displayed as a tree, here termed "epogram" (from "electrostatic potential diagram"). The left side shows an "unrooted" tree. "Unrooted" means that the relatedness of the leaf nodes are displayed without making assumptions about evolutionary timeline. Only the topology of the relatedness is displayed. The "rooted" tree (or phenogram) on the right hand side suggests a common relatedness which generally does not hold. The tree nodes are connected to other nodes by a horizontal and then a vertical line. Please note that, in general, in rooted tree-like diagrams, it is possible to change the order in which sequences that are being referred to as more closely related to each other than others, without altering the meaning of the statement.

It is thus recommended to refer to enzyme's relatedness from the "unrooted" tree diagram. From the epogram one can estimate the kinetic parameter of interest of the query enzyme to be sufficiently similar or close to that of its nearest neighbour enzymes. The length of the tree branches in the epograms does not correspond to a quantitative measure of relatedness but is scaled for optimal display. Examples of PIPSA tree-like diagrams ("epograms") can be found at http://mcm.h-its.org/mcmsoft/pipsa/2.0/examples.html.

Limitations of current implementation in SYCAMORE:

1. Single protein chain modelling only
2. Neglect of protein flexibility
3. Lack of user intervention in:
   1. calculation of electrostatic potentials
   2. PIPSA analysis