1.   N. Blomberg, R.R. Gabdoulline, M. Nilges and R. C. Wade.
Classification of protein sequences by homology modeling and quantitative analysis of electrostatic similarity.
Proteins: Str., Function and Genetics 1999, 37: 379-387.

Protein electrostatics plays a key role in ligand binding and protein protein interactions. Therefore, similarities or dissimilarities in electrostatic potentials can be used as indicators of similarities or dissimilarities in protein function. We here describe a method to compare the electrostatic properties within protein families objectively and quantitatively. Three-dimensional structures are built from database sequences by comparative modeling. Molecular potentials are then computed for these with a continuum solvation model by finite difference solution of the Poisson-Boltzmann equation or analytically as a multipole expansion that permits rapid comparison of very large datasets. This approach is applied to 104 members of the Pleckstrin homology (PH) domain family. The deviation of the potentials of the homology models from those of the corresponding experimental structures is comparable to the variation of the potential in an ensemble of structures from nuclear magnetic resonance data or between snapshots from a molecular dynamics simulation. For this dataset, the results for analysis of the full electrostatic potential and the analysis using only monopole and dipole terms are very similar. The electrostatic properties of the PH domains are generally conserved despite the extreme sequence divergence in this family. Notable exceptions from this conservation are seen for PH domains linked to a Dbl homology (DH) domain and in proteins with internal PH domain repeats.

Blue copper proteins are type-I copper-containing redox proteins whose role is to shuttle electrons from an electron donor to an electron acceptor in bacteria and plants. A large amount of experimental data is available on blue copper proteins; however, their functional characterization is hindered by the complexity of redox processes in biological systems. We describe here the application of a semiquantitative method based on a comparative analysis of molecular interaction fields to gain insights into the recognition properties of blue copper proteins. Molecular electrostatic and hydrophobic potentials were computed and compared for a set of 33 experimentally-determined structures of proteins from seven blue copper subfamilies, and the results were quantified by means of similarity indices. The analysis provides a classification of the blue copper proteins and shows that (I) comparison of the molecular electrostatic potentials provides useful information complementary to that highlighted by sequence analysis; (2) similarities in recognition properties can be detected for proteins belonging to different subfamilies, such as amicyanins and pseudoazurins, that may be isofunctional proteins; (3) dissimilarities in interaction properties, consistent with experimentally different binding specificities, may be observed between proteins belonging to the same subfamily, such as cyanobacterial and eukaryotic plastocyanins; (4) proteins with low sequence identity, such as azurins and pseudoazurins, can have sufficient similarity to bind to similar electron donors and acceptors while having different binding specificity profiles.

We describe how similarity indices may be applied to semiquantitatively compare the molecular interaction fields, such as electrostatic and hydrophobic potentials, of proteins. We then review ways in which this approach has been used to investigate protein interaction properties in diverse applications ranging from in-depth comparison of pairs of proteins to large-scale analysis of hundreds of modelled protein structures.

The oxidation of cytochrome f by the soluble cupredoxin plastocyanin is a central reaction in the photosynthetic electron transfer chain of all oxygenic organisms. Here, two different computational approaches are used to gain new insights into the role of molecular recognition and protein-protein association processes in this redox reaction. First, a comparative analysis of the computed molecular electrostatic potentials of seven single and multiple point mutants of spinach plastocyanin (D42N, E43K, E43N, E43Q/D44N, E59K/E60Q, E59K/E60Q/E43N, Q88E) and the wt protein was carried out. The experimentally determined relative rates (k2) for the set of plastocyanin mutants are found to correlate well (r2   0.90   0.97) with the computed measure of the similarity of the plastocyanin electrostatic potentials. Second, the effects on the plastocyanin/cytochrome f association rate of these mutations in the plastocyanin  eastern site  were evaluated by simulating the association of the wild type and mutant plastocyanins with cytochrome f by Brownian dynamics. Good agreement between the computed and experimental relative rates (k2) (r2   0.89   0.92) was achieved for the plastocyanin mutants. The results obtained by applying both computational techniques provide support for the fundamental role of the acidic residues at the plastocyanin eastern site in the association with cytochrome f and in the overall electron-transfer process.

The E2 enzymes are key enzymes in the ubiquitin and ubiquitin-like protein ligation pathways. To understand the functionality of the different E2 enzymes, we analyzed 190 protein sequences and 211 structures and electrostatic potentials. Key findings include: The ScUbc1 orthologs are defined by a C-terminal UBA domain. An N-terminal sequence motif that is highly conserved in all E2s except for Cdc34 orthologs is important for the stabilization of the L7 loop and is likely to be involved in E1 binding. ScUbc11p has a different electrostatic potential from E2-Cp and other proteins with which it has high sequence similarity but different functionality. All the E2s known to ubiquitinate histones have a negative potential. The members of the NCUBE family have a positive electrostatic potential, although its form is different from that of the SUMO conjugating E2s. The specificities of only the ScUbc4/Ubc5 and ScUbc1p orthologs are reflected in their L4 and L7 loops.

WW domains are small globular protein interaction modules found in a wide spectrum of proteins. They recognize their target proteins by binding specifically to short linear peptide motifs that are often proline-rich. To infer the determinants of the ligand binding propensities of WW domains, we analyzed 42 WW domains. We built models of the 3D structures of the WW domains and their peptide complexes by comparative modeling supplemented with experimental data from peptide library screens. The models provide new insights into the orientation and position of the peptide in structures of WW domain-peptide complexes that have not yet been determined experimentally. From a protein interaction property similarity analysis (PIPSA) of the WW domain structures, we show that electrostatic potential is a distinguishing feature of WW domains and we propose a structure-based classification of WW domains that expands the existent ligand-based classification scheme. Application of the comparative molecular field analysis (CoMFA), grid/golpe and comparative binding energy (COMBINE) analysis methods permitted the derivation of quantitative structure-activity relationships (QSARs) that aid in identifying the specificity-determining residues within WW domains and their ligand-recognition motifs. Using these QSARs, a new group-specific sequence feature of WW domains that target arginine-containing peptides was identified. Finally, the QSAR models were applied to the design of a peptide to bind with greater affinity than the known binding peptide sequences of the yRSP5-1 WW domain. The prediction was verified experimentally, providing validation of the QSAR models and demonstrating the possibility of rationally improving peptide affinity for WW domains. The QSAR models may also be applied to the prediction of the specificity of WW domains with uncharacterized ligand-binding properties.