Volume 24, Issue 9 pp. 1486-1494

Article

Free Access

Linking structural features of protein complexes and biological function

Gopichandran Sowmya,

Gopichandran Sowmya

Department of Chemistry and Biomolecular Sciences, Macquarie University, Sydney, New South Wales 2109, Australia

Search for more papers by this author

Edmond J. Breen,

Edmond J. Breen

Australian Proteome Analysis Facility (APAF), Macquarie University, Sydney, New South Wales 2109, Australia

Search for more papers by this author

Shoba Ranganathan,

Corresponding Author

Shoba Ranganathan

Department of Chemistry and Biomolecular Sciences, Macquarie University, Sydney, New South Wales 2109, Australia

Correspondence to: Shoba Ranganathan, Department of Chemistry and Biomolecular Sciences, Macquarie University, Sydney, New South Wales 2109, Australia. E-mail: [email protected]Search for more papers by this author

Gopichandran Sowmya,

Gopichandran Sowmya

Department of Chemistry and Biomolecular Sciences, Macquarie University, Sydney, New South Wales 2109, Australia

Search for more papers by this author

Edmond J. Breen,

Edmond J. Breen

Australian Proteome Analysis Facility (APAF), Macquarie University, Sydney, New South Wales 2109, Australia

Search for more papers by this author

Shoba Ranganathan,

Corresponding Author

Shoba Ranganathan

Department of Chemistry and Biomolecular Sciences, Macquarie University, Sydney, New South Wales 2109, Australia

First published: 30 June 2015

https://doi.org/10.1002/pro.2736

Citations: 34

Share a link

Email
Wechat
Bluesky

Abstract

Protein–protein interaction (PPI) establishes the central basis for complex cellular networks in a biological cell. Association of proteins with other proteins occurs at varying affinities, yet with a high degree of specificity. PPIs lead to diverse functionality such as catalysis, regulation, signaling, immunity, and inhibition, playing a crucial role in functional genomics. The molecular principle of such interactions is often elusive in nature. Therefore, a comprehensive analysis of known protein complexes from the Protein Data Bank (PDB) is essential for the characterization of structural interface features to determine structure–function relationship. Thus, we analyzed a nonredundant dataset of 278 heterodimer protein complexes, categorized into major functional classes, for distinguishing features. Interestingly, our analysis has identified five key features (interface area, interface polar residue abundance, hydrogen bonds, solvation free energy gain from interface formation, and binding energy) that are discriminatory among the functional classes using Kruskal-Wallis rank sum test. Significant correlations between these PPI interface features amongst functional categories are also documented. Salt bridges correlate with interface area in regulator-inhibitors (r = 0.75). These representative features have implications for the prediction of potential function of novel protein complexes. The results provide molecular insights for better understanding of PPIs and their relation to biological functions.

Introduction

Protein–protein interaction (PPI) is critical for molecular functions in living systems. PPIs are associated with catalysis, regulation, signaling, immunity, and inhibition, thereby playing a critical role in functional genomics.1 Extensive studies on the existing PPI complexes are the key to understanding cellular machinery as reviewed elsewhere.1-5 With modern experimental techniques such as two hybrid systems, protein fragment complementation, tandem affinity purification methods and protein arrays, several interacting protein pairs have been detected in large-scale studies,6 although their biological role may not be well characterized or known. Comprehensive homology modeling techniques of known interacting proteins combined with docking studies and PPI data helps in understanding structural assembly for functional preferences as shown for integrin αvβ6 heterodimer (two different protein subunits) complex.7 Thus, the analysis of PPIs is essential in predicting important biological functions.8, 9 It is well known that structure-based characterization of multimeric proteins is the key to ascribing biological functional annotation.10

The availability of protein–protein complexes at the Protein Data Bank (PDB)11 has enabled sequence and structural investigations on the existing complexes to decipher their recognition mechanisms and binding principles for decades.3, 12-15 The classical work by Chothia and Janin16 with three protein complexes defined ‘hydrophobicity’ (nonpolar interactions) as the major stabilizing factor in PPI, which has been affirmed with larger datasets of protein complexes.12, 14, 17-19 Conversely, shape complementarity, polar interactions, hydrogen bonding, and salt bridges are believed to primarily contribute to binding specificity and free energy of binding.20-23 Furthermore, a number of physicochemical factors known to govern protein–protein association include interface size, planarity, sphericity, complementarity, types of amino acid chemical groups, hydrophobicity, electrostatic interactions, H-bonds, distribution of binding energy, sequence conservation, and conserved residue clusters.12, 16, 21, 24-27 However, extensive surveys carried out thus far by various groups' typically average structural features over diverse datasets of protein–protein complexes, obscuring information on individual proteins' structural integrity. In an earlier study, we developed a homology model for integrin αvβ6 heterodimer7 where we found that the subunit interface in our model as well as the template X-ray structures clearly showed an increase of polar residues compared with the surface or the complex. Therefore, it is essential to revisit interface analysis to understand PPI binding principles, using a large nonredundant structural dataset.

The classification of protein–protein complexes based on their composition, affinity, interface stability, and lifetime association into different groups has gained momentum in the past decade,28-30 although the boundaries between these classes is often indefinite, based on physiological conditions.31 Alternatively, the classification of complexes into major functional groups can be valuable in relating structural data to biological functions for better understanding of PPIs.32 Moreover, the classification of protein–protein complexes based on functions and their usefulness in improving prediction accuracy has been observed recently.33 Therefore, for analyzing PPI primarily based on functions, a nonredundant set of complexes, with structural information, is essential.

Nonidentical protein subunits (or chains) noncovalently interact to form ‘heteromers’ with diverse functionality. Heterodimeric interactions are commonly found in enzyme-inhibitors, enzyme complexes, antibody–antigen, signal proteins and cell cycle proteins, and also include transient complexes. Dimeric interactions are amongst the strongest and most extensive in nature.12, 14 In this study, we have created a nonredundant dataset of 278 heteromeric protein structure dimers from the PDB, for characterization of their interface structural features to determine whether biochemical function is related to interface features. Several physico–chemical features such as interface size, interface area, hydrogen bonds (H-bonds), salt bridges, solvation free energy gain (ΔⁱG), binding energy (BE), and interface electrostatic energy (ΔΔG_el) were investigated to study discriminatory features prevailing in different functional groups. Our analysis has identified key features [interface area, interface polarity abundance (P% − NP%), H-bonds, ΔⁱG, and BE] that are significantly different between the functional groups. This result has implications for function prediction for orphan proteins, where interacting partners are known and heteromeric complexes can be structurally modeled with high confidence.

Results and Discussion

Our nonredundant dataset (described in Materials and Methods: Heterodimer dataset) comprised 278 complexes (Table 1). Our previous dataset is 98.7% of the current dataset with only five new entries in the current dataset (underlined in Table 1). The dataset was then grouped into important functional groups and protein types to better understand PPIs among these divisions of complexes. The dataset comprises enzymes (40), regulators (144), enzyme-inhibitors (25), regulator-inhibitors (27) and immune (18) and biological assembly (24) complexes, of which 247 are globular and 31 membranous. The distribution of complexes under different functional classes is unbiased. Enzyme complexes are formed when two enzyme subunits interact to achieve a ‘catalytic’ function. For example, PDB ID: 1EUD succinyl coenzyme A (succinyl-CoA) synthetase is formed by the interaction between succinyl CoA synthetase α chain (311 AA) and succinyl-CoA synthetase β chain (396 AA). The succinyl-CoA synthetase (SCS) protein catalyses a reversible conversion of succinyl-CoA and succinate, coupled with phosphorylation/dephosphorylation. Enzyme complexes considered in this study do not include enzyme-inhibitor complexes. Enzyme-inhibitor complexes are formed when an enzyme and an inhibitor protein interact to achieve an ‘inhibitory’ function. For example, PDB ID: 1ARO is T7 RNA polymerase (883 AA) complexed to the transcriptional inhibitor T7 lysozyme (151 AA). The lysozyme binds at a site distant from the polymerase active site, suggesting an indirect mechanism of inhibition.

Table 1. A Dataset of Heterodimer Protein Complexes (278) Divided into Literature Driven Functional Groups with Major Protein Types. New entries added to the earlier dataset [42] are underlined.

Functional groups	Globular (247)							Membrane (31)
Enzymes (40)	1EUD	2APO	2O2V	3EP6	3O3O	3REQ	4HST	3RGW
	1GK9	2BLF	2QKL	3FGR	4EMJ	3ZVQ	4MRT
	1NME	2CG5	3A8G	3GA9	3L91	4BMP	4NFU
	1VRA	2GA9	3AON	3JTQ	3M7F	4DBG
	1KA9	2XFG	3AYH	3DSS	3ML1	4GED
	1SVD	2ZIV	3DRA	3MWD	3R07	4HNX
Regulators (144)	1JEQ	2G2S	3B0C	3LF4	4AT7	4G6T	4IYP	1FS0
	1JKG	2H9A	3B0Z	3MCB	4AWX	3VRD	4JHP	1H32
	1LSH	2HRK	3CPT	3MKR	4BJJ	3W9C	4KBM	1NRJ
	1N1J	2IW5	3CX8	3MXN	3O2P	3ZNZ	4KT1	1Z0J
	1OF5	2VDB	3DGP	3NW0	3OG6	3ZYI		1Z5Y
	1OO0	2P1M	3DPL	3GC3	3OJM	4BL7		2GSK
	1ORY	2QSF	3EGV	3HHM	3SBT	4C2A		2PQN
	1US7	2QWO	3CLS	3KLD	3THO	4DVG		2V6X
	1WQJ	2V3B	3D3B	3KYJ	3TU3	4GDX		3AQF
	1YKH	2VN6	3EI3	3MCA	3VF0	4JE3		3BS5
	2D74	2WD5	3F6Q	3N1M	4C9B	4JEH		4CXF
	1DJ7	2ZFD	3FMO	3NV0	4CBU	4KHA		4G1M
	1H2V	2ZSI	3FQD	3OSS	4CGY	4KMO		1ZHH
	1USU	3ANW	3GB8	3PGE	4CT0	4L2I		2FCW
	1WMH	3AWU	3F62	3Q87	4DEY	4M69		2V8S
	1Z3E	3AXJ	3H7H	3T5X	4E4W	4M6W		3MP7
	1ZBX	3BEG	3HZH	3TBI	4EGC	4NQW		3NY7
	2CKL	3A2F	3IF8	3VZ9	4ETP	4O8Y		3OUN
	2DYO	3ABE	3K8P	3ZET	4EUK	4FZV		3V8X
	2FH5	3AU4	3KXC	4A5U	4EYY	4GQ2		4G7X
Enzyme-inhibitors (25)	1ARO	1AY7	2OZN	3QN1	3N4I	4DRI	4LV5
	1R0R	1WPX	3CKI	2O3B	3SHG	4F9C
	1R8O	1XG2	3DBO	2VLQ	4F6U	4HT3
	1UGH	2B42	3IEY	3R24	3FJU	4I1S
Regulator-inhibitors (27)	1A22	2FOM	2QC1	3NVN	3OQ3	4BI8	4GVB	1E44
	1M1E	2HDI	3N40	3ONA	3QQ8	4GAF		1T6B
	1JMA	2F4M	2RAW	3AA7	3NYB	4B8A
	1XOU	2OMZ	2VSM	3BTP	3OJA	4G94
Immune complexes (18)	1Z92	2PTT	3DLQ	3PV6	4HFF	4KT3		1T0P
	2FHZ	2Z64	3FPU	3RNQ	4J38			1V74
	2P45	3DI3	3MJ7	3WA5	4K12
Biological assembly (24)	1XEW	2Z5B	3K1R	3LQC	3W8I	4H4K		1GL4
	2FTX	3CQC	3KCP	3VYR	4AP2	4HPL		1NPE
	2PA8	3FPN	3KF6	3VU9	4F48	4IU2		3A4U
								3C5X
								3LBX
								3M1C

Regulatory complexes are formed between two protein subunits to achieve a ‘regulatory’ or ‘cellular’ function. For example, PDB ID: 1JEQ is a Ku heterodimer, formed by Ku70 (609 AA) and Ku80 (565 AA) subunits, contributing to genomic integrity by binding to DNA double-strand breaks and enabling repair by nonhomologous end-joining pathway. The regulatory complexes do not include regulator-inhibitor complexes, which are formed between a regulatory protein subunit and an inhibitor protein to achieve an ‘inhibitory’ function. For example, PDB ID: 1A22 is a complex between the G120R mutant (191 AA) of the human growth hormone and its cognate receptor, the human growth hormone receptor (238 AA). The G120R mutant of human growth hormone (hGH) is an antagonist binding to the growth hormone receptor. An example for each of these functional groups is shown in Figure 1.

Details are in the caption following the image — **Figure 1**
Open in figure viewer PowerPoint

Examples of protein–protein complex structures from each functional group. (a) An enzyme complex (PDB: 1EUD) formed between succinyl-CoA synthetase α chain and succinyl-CoA synthetase β chain is shown in yellow and green colors, respectively. (b) TACE-N-TIMP-3 enzyme-inhibitor complex (PDB: 3CKI) with ADAM (a disintegrin and metalloproteinase) inhibition by TIMP-3 (tissue inhibitor of metalloproteinases 3) is shown in light brown and red, respectively. (c) An immune complex (PDB: 2PTT) between NK cell receptor 2B4 (CD244) bound to CD48 is shown in black and dark brown, respectively. (d) A regulatory SoxAX protein (involved in enzymatic oxidation of thiosulfate; PDB:1H32) of the diheme cytochrome C (in cyan colored) and cytochrome C (in light green) is shown. (e) A regulator–inhibitor (PDB:1E44) of the cytotoxic domain of colicin E3 in complex with its immunity protein is shown in green and red, respectively. (f) A biological assembly of the nuclear pore complex (NPC) between Nup107 and Nup133 (PDB: 3CQC) is shown in cyan and dark blue, respectively.

PPI features among functional groups

Of all the physicochemical features described in the Materials and Methods section, five were significant in discriminating between the functional categories. The results from these features, measured in terms of six parameters (Fig. 2), using a Kruskal-Wallis rank sum test (P-values in Table 2) are presented below, along with sets of correlated properties.

Table 2. Discriminatory PPI Features Among Functional Groups (Using Kruskal-Wallis Rank Sum Test)

Functional groups	P-value	Q-value
Solvent free energy gain (ΔⁱG)	7.08E−06	0.000128
Interface area	2.25E−05	0.000382
Binding energy	3.11E−05	0.000497
Interface polarity abundance (P%−NP%)	4.25E−05	0.000638
Hydrogen bonds	7.54E−04	0.009964
Solvent free energy gain (ΔⁱG) P-value	7.12E−04	0.009964

Abbreviations: ΔⁱG, solvent free energy gain; ΔASA, interface area; ΔΔG_el, interface electrostatic free energy calculation; BE, binding energy; H-bonds, hydrogen bonds; ICM, internal coordinate mechanics; NP, nonpolar residues; PDB, Protein Data Bank; P, polar residues; PPI, protein–protein interaction.

Interface area among complexes

The standard value reported by Lo Conte and colleagues for a protein interface is 1600 Å² (±400).26 Bahadur and colleagues showed that the range of interface area (ΔASA) extends from 500 to 7000 Å², with a mean of 1970 Å²,34 while Caffrey et al. showed that interface area ranges from 415 to 2361 Å² for heterodimer complexes.35 Our analysis (described in Materials and Methods: Interface analysis) shows that the interface area per subunit (B/2) ranges from ∼350 to 9500 Å² (±1100) and differs for each functional group as shown in Figure 2. The interface area is significantly different among functional groups with P = 2.25E−05 (Table 2).

Interface polarity abundance

The difference in percentages between interface polar residues and nonpolar residues (P% − NP%) gives the measure of the abundance of polar or nonpolar residues at the interface.32 The interface polarity abundance (P% − NP%) measure (described in Materials and Methods: Interface analysis) is significantly different among the different functional categories with P = 4.25E−05 (Table 2 and Fig. 2).

Hydrogen bonds among complexes

The stability of protein–protein binding depends on the number of hydrogen bonds (H-bonds) and salt bridges formed between the two interacting subunits. On an average, 10.1 H-bonds are formed at a protein–protein interface, with one H-bond per 170 Å² interface area with an r value of 0.84 observed between H-bonds and interface area.26 The r value between H-bonds and interface area calculated using different dataset size and nature of data varies from 0.75 to 0.89,12, 14, 21, 26, 36, 37 with an average of 0.24 H-bonds per interface residue in heterodimers. High H-bond density per interface residue (0.64) with dominant charged and hydrophilic/polar residues at the heterodimer protein interfaces is also demonstrated.37 Our statistical analysis (described in Materials and Methods: Intermolecular hydrogen bonds calculation) shows that the total number of intermolecular H-bonds for each functional group is significantly different with P = 7.54E−04 (Table 2 and Fig. 2).

Solvent free energy gain upon interface formation

The solvation free energy gain upon interface formation (ΔⁱG)) [calculated as described in Materials and Methods: calculation of solvation free energy gain upon interface formation (ΔⁱG)] is significantly different among functional groups with P = 7.08E−06 04 (Table 2) as shown in Figure 2.

Binding energy at the interface

To study the strengths of binding among functional groups, we estimated the binding free energy (BE; also called binding affinity calculated as described in Materials and Methods: Binding energy calculation) of the dataset. The BEs are significantly different among functional groups with P = 3.11E−05 (Table 2) as shown in Figure 2.

Correlations amongst interface physiochemical features

Interface polarity abundance (P% − NP%) shows limited correlation with charged residues at the interface (r = 0.61) for the heterodimer dataset. However, this is not unexpected as charged residues are included in the polar residue set. Also, the protein complexes in our dataset show high correlation between intermolecular H-bonds and interface area (r = 0.90) as previously observed.12, 14, 21, 26, 36, 37 Moreover, H-bonds are correlated to BE (r = −0.70). Salt bridges (described in Materials and Methods: Intermolecular salt bridges calculation) across the interface do not show any significant correlation to all other features, which is in accord with another study.21 Although this parameter was not statistically significant between the different functional groups, it is correlated to interface area in regulator-inhibitor complexes (r = 0.75; see Supporting Information Fig. S1). This shows that intermolecular salt bridges are an important structural feature in some functional complexes.

The solvation free energy gain upon interface formation (ΔⁱG) shows high correlation with interface area in the heterodimer dataset (r = −0.88), suggesting ΔⁱG is an important feature in characterizing PPIs. Interestingly, biological assembly and immune complexes showed the least correlation of ΔⁱG with interface area (r = −0.67) as compared with other functional groups as shown in Supporting Information Figure S2. The ΔⁱG and BE are also highly correlated among the dataset (r = 0.88). Our analysis shows a high correlation between BE and interface area in the dataset (r = 0.96) as previously observed.38

The electrostatic component of binding free energy (ΔΔG_el) was studied to quantify the electrostatic free energy favoring protein–protein interaction among functional groups of complexes (described in Materials and Methods: Interface electrostatic free energy calculation (ΔΔG_el)). Interface electrostatic energy component of BE (Supporting Information Fig. S3) shows distribution of charges in the dataset with electrostatic energy contributing to destabilizing PPIs in a few complexes while stabilizing PPIs in the others, suggesting that quantification of accurate interface electrostatic component contributing to BE is often a nontrivial task. The frequency distribution also shows similar trends for interface electrostatic energy (Supporting Information Fig. S4). Although ΔΔG_el values show limited correlation with interface area (r = −0.47), the enzymes complexes show correlation between ΔΔG_el and interface area (r = −0.61), as opposed to other groups (r < 0.5).

These discriminatory PPI features hold significantly different among functional groups in globular proteins with P < 0.05 as shown in supplementary Supporting Information Table S1. Since there is insufficient information regarding all the functional groups in membrane proteins, the discriminatory features among functional groups in membrane proteins is not clear at this point. These observations corroborate the need for classification of complexes in determining their combinatorial features and drawing consensus for common patterns in protein–protein recognition. Incorporation of these combinatorial features among protein functional groups is necessary to develop models for residue-level protein–protein binding prediction and analysis, and also in utilizing PPI information for the prediction of potential protein functions in future studies.

Materials and Methods

Heterodimer dataset

We created a nonredundant protein heterodimer (comprising two different protein subunits) dataset, the 3D structures of which were determined by X-ray crystallography, from the PDB, using the RCSB's advanced search interface. The search criteria were: (i) resolution <= 3Å and (ii) protein size > 50 residues, as described in earlier studies (iii) limited to “experimental data” to obtain high resolution true structures and avoid short peptides, synthetic, or artificial complexes and (iv) the number of chains, entities, and oligomeric state is set at two to obtain dimers with two unique or different chains, (v) exclude DNA or RNA or a hybrid of such molecules with proteins or otherwise and (vi) sequence identity cut-off is set to 30% and (vii) the select parameter was set so that entries with mutations were not included in the dataset. The redundancy among heterodimer complexes was further removed using the USEARCH program39 at sequence identity cut-off of 20%, as this threshold eliminates remote homology of 25%40 seen in structures as well. The criteria set out are comparable with those of Janin and coworkers,41 who have used a resolution cutoff for X-ray structures (3.25Å) and chain length (minimum of 30 residues) for defining a benchmark dataset for complexes with experimental binding energies, toward developing a method for binding energy prediction.

Grouping based on function and protein type

The assembly of proteins into functional complexes is essential in biology. Therefore, the characterization of these functional complexes is an essential step in deciphering their binding principles. Hence, we have grouped the dataset into major functional groups such as enzymes (E), enzyme-inhibitors (EI), regulators (R), regulator-inhibitors (RI), immune (I), and biological assembly (BA) complexes, as described in the PDB header. In a few cases, where the complexes had more than one function, the functional group assigned is based on their primary role obtained from literature. The dataset comprised globular (247) and membranous (31) complexes.

Structural Analysis

Interface analysis

A protein–protein interface is identified by calculating the change in solvent accessible surface area (ΔASA) upon binding.42 Surface Racer 5.0 program43 with a probe radius of 1.4Å and Lee and Richards implementation42 was used to calculate ASA. ΔASA (interface area) of a heterodimer complex was calculated as shown in Eq. 1.

$urn:x-wiley:09618368:media:pro2736:pro2736-math-0001$ (1)

The identified interface residues were further filtered based on the criteria that their ΔASA > 0.1Å².44 The amount of polar, nonpolar, and charged residues at the interface was then estimated for the dataset. Polar residues considered in the analysis are R, N, D, E, Q, H, K, S, T, and Y.45 We have computed the relative abundance of polar or nonpolar residues [“interface polarity abundance (P% − NP%)”] compared with nonpolar residues at the interface as the difference between the percentage of interface polar residues and the percentage of interface nonpolar residues.

Intermolecular hydrogen bonds calculation

The hydrogen atoms covalently bound between two electronegative atoms and contributing to electrostatics was calculated using HBPLUS program.46 The output file of HBPLUS contains information on all donor and acceptor atoms, angles and distances within the distance of 4 Å and then filtered for intermolecular hydrogen bonds, where the hydrogen bond donor and acceptor atoms are from two different subunits.

Intermolecular salt bridges calculation

The salt bridges formed between two oppositely charged side-chain atoms, (i.e., basic and acidic amino acids) within a distance of 4 Å and contributing to the stability and electrostatics of the protein complex was calculated using the SBION program,47 and intermolecular salt bridges were then extracted such that the oppositely charged atoms are from two different subunits.

Calculation of solvation free energy gain upon interface formation (ΔⁱG)

The ΔⁱG of protein complex arises from the change in solvation energy as well as contact-dependent and electrostatic interactions of the subunits, quantifying the solvation free energy gained upon interface formation. The difference in total solvation energies between the isolated and complexed structure gives the solvation free energy gain at the interface (in kcal/mol), The PDBePISA webserver48 was used to calculate ΔⁱG for the heterodimer dataset.

Binding energy calculation (BE)

The interaction between two protein subunits can be characterized in terms of binding free energy. The BE term (comprising electrostatic, van der Waals, hydrophobic, and entropic terms) gives an indication of strong and weak intermolecular forces, with the most negative value considered the strongest. The binding free energy of 278 protein complexes was calculated using the DCOMPLEX program,49 which uses DFIRE-based potentials.50 DCOMPLEX is the most widely used program for calculating binding energies of protein complexes as it has been benchmarked to reproduce experimental binding free energy values.

Interface electrostatic free energy (ΔΔG_el) calculation

Interface electrostatic free energy component of the binding free energy (ΔΔG_el) was calculated as the difference in electrostatic free energies of the complex and of the free protein subunits as follows:

$urn:x-wiley:09618368:media:pro2736:pro2736-math-0002$ (2)

where ΔG_el(A:B), ΔG_el(A), and ΔG_el(B) are the electrostatic energies of the complex AB, monomer A, and monomer B as described previously.51 ICM software52 was used to calculate the electrostatic free energies of the complex and the free subunits, as DCOMPLEX does not provide a breakup of the binding energy into components. The ICM REBEL (Rapid Exact-Boundary Electrostatics)53 module included in the ICM-Pro package uses ‘boundary element’ method to solve Poisson equation for the protein with analytical molecular surface as dielectric boundary. The ICM method is fast and accurate and estimates the electrostatic energy of proteins surrounded by continuous aqueous solution. The energy solved by this method consists of solvation energy and coulomb energy.

Statistical analysis

The Kruskal-Wallis rank sum test, a nonparametric method,54 is used to test whether the mean ranks for the PPI features in all functional groups are the same. The discriminatory PPI features among functional groups were thus tested for statistical significance with P < 0.05 (for the Kruskal-Wallis rank sum test) in RStudio.55 Spearman's rank correlation coefficient has been used to measure of statistical dependence between the interface features, since it assesses how well the relationship between two properties can be described using a monotonic function.56

Acknowledgments

G.S. is the recipient of an iMQRES PhD scholarship from Macquarie University. The authors are grateful to Prof Pandjassarame Kangueane and Prof. Michael Gromiha for valuable discussions.

Supporting Information

References

1Sowmya G, Ranganathan S (2014) Protein-protein interactions and prediction: a comprehensive overview. Protein Pept Lett 21: 779–789.
10.2174/09298665113209990056
CAS PubMed Web of Science® Google Scholar
2Reichmann D, Rahat O, Cohen M, Neuvirth H, Schreiber G (2007) The molecular architecture of protein-protein binding sites. Curr Opin Struct Biol 17: 67–76.
10.1016/j.sbi.2007.01.004
CAS PubMed Web of Science® Google Scholar
3Keskin O, Gursoy A, Ma B, Nussinov R (2008) Principles of protein-protein interactions: what are the preferred ways for proteins to interact? Chem Rev 108: 1225–1244.
10.1021/cr040409x
CAS PubMed Web of Science® Google Scholar
4Sudha G, Nussinov R, Srinivasan N (2014) An overview of recent advances in structural bioinformatics of protein-protein interactions and a guide to their principles. Prog Biophys Mol Biol 116:141–150.
10.1016/j.pbiomolbio.2014.07.004
PubMed Web of Science® Google Scholar
5Bahadur RP, Zacharias M (2008) The interface of protein-protein complexes: analysis of contacts and prediction of interactions. Cell Mol Life Sci 65: 1059–1072.
10.1007/s00018-007-7451-x
CAS PubMed Web of Science® Google Scholar
6Piehler J (2005) New methodologies for measuring protein interactions in vivo and in vitro. Curr Opin Struct Biol 15: 4–14.
10.1016/j.sbi.2005.01.008
CAS PubMed Web of Science® Google Scholar
7Sowmya G, Khan JM, Anand S, Ahn SB, Baker MS, Ranganathan S (2014) A site for direct integrin αvβ6.uPAR interaction from structural modelling and docking. J Struct Biol 185: 327–335.
10.1016/j.jsb.2014.01.001
CAS PubMed Web of Science® Google Scholar
8Letovsky S, Kasif S (2003) Predicting protein function from protein/protein interaction data: a probabilistic approach. Bioinformatics 19 Suppl 1: i197–204.
10.1093/bioinformatics/btg1026
PubMed Web of Science® Google Scholar
9Nabieva E, Jim K, Agarwal A, Chazelle B, Singh M (2005) Whole-proteome prediction of protein function via graph-theoretic analysis of interaction maps. Bioinformatics 21 (Suppl 1): i302–i310.
10.1093/bioinformatics/bti1054
CAS PubMed Web of Science® Google Scholar
10Wiederstein M, Gruber M, Frank K, Melo F, Sippl MJ (2014) Structure-based characterization of multiprotein complexes. Structure 22: 1063–1070.
Google Scholar
11Berman HM, Westbrook J, Feng Z, Gilliland G, Bhat TN, Weissig H, Shindyalov IN, Bourne PE (2000) The protein data bank. Nucleic Acids Res 28: 235–242.
10.1093/nar/28.1.235
CAS PubMed Web of Science® Google Scholar
12Jones S, Thornton JM (1995) Protein-protein interactions: a review of protein dimer structures. Prog Biophys Mol Biol 63: 31–65.
10.1016/0079-6107(94)00008-W
CAS PubMed Web of Science® Google Scholar
13Gromiha MM, Yokota K, Fukui K (2009) Energy based approach for understanding the recognition mechanism in protein-protein complexes. Mol Biosyst 5: 1779–1786.
10.1039/b904161n
CAS PubMed Web of Science® Google Scholar
14Jones S, Thornton JM (1996) Principles of protein-protein interactions. Proc Natl Acad Sci USA 93: 13–20.
10.1073/pnas.93.1.13
CAS PubMed Web of Science® Google Scholar
15Gilson MK, Zhou HX (2007) Calculation of protein-ligand binding affinities. Annu Rev Biophys Biomol Struct 36: 21–42.
10.1146/annurev.biophys.36.040306.132550
CAS PubMed Web of Science® Google Scholar
16Chothia C, Janin J (1975) Principles of protein-protein recognition. Nature 256: 705–708.
10.1038/256705a0
CAS PubMed Web of Science® Google Scholar
17Miller S, Lesk AM, Janin J, Chothia C (1987) The accessible surface area and stability of oligomeric proteins. Nature 328: 834–836.
10.1038/328834a0
CAS PubMed Web of Science® Google Scholar
18Bahadur RP, Chakrabarti P, Rodier F, Janin J (2003) Dissecting subunit interfaces in homodimeric proteins. Proteins 53: 708–719.
10.1002/prot.10461
CAS PubMed Web of Science® Google Scholar
19Levy ED (2010) A simple definition of structural regions in proteins and its use in analyzing interface evolution. J Mol Biol 403: 660–670.
10.1016/j.jmb.2010.09.028
CAS PubMed Web of Science® Google Scholar
20McCoy AJ, Chandana Epa V, Colman PM (1997) Electrostatic complementarity at protein/protein interfaces. J Mol Biol 268: 570–584.
10.1006/jmbi.1997.0987
CAS PubMed Web of Science® Google Scholar
21Xu D, Tsai CJ, Nussinov R (1997) Hydrogen bonds and salt bridges across protein-protein interfaces. Protein Eng 10: 999–1012.
10.1093/protein/10.9.999
CAS PubMed Web of Science® Google Scholar
22Kundrotas PJ, Alexov E (2006) Electrostatic properties of protein-protein complexes. Biophys J 91: 1724–1736.
10.1529/biophysj.106.086025
CAS PubMed Web of Science® Google Scholar
23Wong ET, Na D, Gsponer J (2013) On the importance of polar interactions for complexes containing intrinsically disordered proteins. PLoS Comput Biol 9: e1003192.
10.1371/journal.pcbi.1003192
CAS PubMed Web of Science® Google Scholar
24Saha RP, Bahadur RP, Pal A, Mandal S, Chakrabarti P (2006) ProFace: a server for the analysis of the physicochemical features of protein-protein interfaces. BMC Struct Biol 6: 11
10.1186/1472-6807-6-11
CAS PubMed Web of Science® Google Scholar
25Dong F, Zhou HX (2006) Electrostatic contribution to the binding stability of protein-protein complexes. Proteins 65: 87–102.
10.1002/prot.21070
CAS PubMed Web of Science® Google Scholar
26Lo Conte L, Chothia C, Janin J (1999) The atomic structure of protein-protein recognition sites. J Mol Biol 285: 2177–2198.
10.1006/jmbi.1998.2439
CAS PubMed Web of Science® Google Scholar
27Guharoy M, Chakrabarti P (2010) Conserved residue clusters at protein-protein interfaces and their use in binding site identification. BMC Bioinformatics 11: 286.
10.1186/1471-2105-11-286
CAS PubMed Web of Science® Google Scholar
28Nooren IM, Thornton JM (2003) Diversity of protein-protein interactions. Embo J 22: 3486–3492.
10.1093/emboj/cdg359
CAS PubMed Web of Science® Google Scholar
29Park SH, Reyes JA, Gilbert DR, Kim JW, Kim S (2009) Prediction of protein-protein interaction types using association rule based classification. BMC Bioinformatics 10: 36.
10.1186/1471-2105-10-36
CAS PubMed Web of Science® Google Scholar
30Nooren IM, Thornton JM (2003) Structural characterisation and functional significance of transient protein-protein interactions. J Mol Biol 325: 991–1018.
10.1016/S0022-2836(02)01281-0
CAS PubMed Web of Science® Google Scholar
31Ozbabacan SE, Engin HB, Gursoy A, Keskin O (2011) Transient protein-protein interactions. Protein Eng Des Sel 24: 635–648.
10.1093/protein/gzr025
CAS PubMed Web of Science® Google Scholar
32Sowmya G, Anita S, Kangueane P (2011) Insights from the structural analysis of protein heterodimer interfaces. Bioinformation 6: 137–143.
10.6026/97320630006137
PubMed Google Scholar
33Yugandhar K, Gromiha MM (2014) Protein-protein binding affinity prediction from amino acid sequence. Bioinformatics 30: 3583–3589.
10.1093/bioinformatics/btu580
CAS PubMed Web of Science® Google Scholar
34Bahadur RP, Chakrabarti P, Rodier F, Janin J (2004) A dissection of specific and non-specific protein-protein interfaces. J Mol Biol 336: 943–955.
10.1016/j.jmb.2003.12.073
CAS PubMed Web of Science® Google Scholar
35Caffrey DR, Somaroo S, Hughes JD, Mintseris J, Huang ES (2004) Are protein-protein interfaces more conserved in sequence than the rest of the protein surface? Protein Sci 13: 190–202.
10.1110/ps.03323604
CAS PubMed Web of Science® Google Scholar
36Zhanhua C, Gan JG, Lei L, Mathura VS, Sakharkar MK, Kangueane P (2005) Identification of critical heterodimer protein interface parameters by multi-dimensional scaling in euclidian space. Front Biosci 10: 844–852.
10.2741/1578
PubMed Google Scholar
37Zhanhua C, Gan JG, Lei L, Sakharkar MK, Kangueane P (2005) Protein subunit interfaces: heterodimers versus homodimers. Bioinformation 1: 28–39.
10.6026/97320630001028
PubMed Google Scholar
38Chen J, Sawyer N, Regan L (2013) Protein-protein interactions: general trends in the relationship between binding affinity and interfacial buried surface area. Protein Sci 22: 510–515.
10.1002/pro.2230
CAS PubMed Web of Science® Google Scholar
39Edgar RC (2010) Search and clustering orders of magnitude faster than BLAST. Bioinformatics 26: 2460–2461.
10.1093/bioinformatics/btq461
CAS PubMed Web of Science® Google Scholar
40Sander C, Schneider R (1991) Database of homology-derived protein structures and the structural meaning of sequence alignment. Proteins 9: 56–68.
10.1002/prot.340090107
CAS PubMed Web of Science® Google Scholar
41Fleishman SJ, Whitehead TA, Strauch EM, Corn JE, Qin S, Zhou HX, Mitchell JC, Demerdash ON, Takeda-Shitaka M, Terashi G, Moal IH, Li X, Bates PA, Zacharias M, Park H, Ko JS, Lee H, Seok C, Bourquard T, Bernauer J, Poupon A, Azé J, Soner S, Ovali SK, Ozbek P, Tal NB, Haliloglu T, Hwang H, Vreven T, Pierce BG, Weng Z, Pérez-Cano L, Pons C, Fernández-Recio J, Jiang F, Yang F, Gong X, Cao L, Xu X, Liu B, Wang P, Li C, Wang C, Robert CH, Guharoy M, Liu S, Huang Y, Li L, Guo D, Chen Y, Xiao Y, London N, Itzhaki Z, Schueler-Furman O, Inbar Y, Potapov V, Cohen M, Schreiber G, Tsuchiya Y, Kanamori E, Standley DM, Nakamura H, Kinoshita K, Driggers CM, Hall RG, Morgan JL, Hsu VL, Zhan J, Yang Y, Zhou Y, Kastritis PL, Bonvin AM, Zhang W, Camacho CJ, Kilambi KP, Sircar A, Gray JJ, Ohue M, Uchikoga N, Matsuzaki Y, Ishida T, Akiyama Y, Khashan R, Bush S, Fouches D, Tropsha A, Esquivel-Rodríguez J, Kihara D, Stranges PB, Jacak R, Kuhlman B, Huang SY, Zou X, Wodak SJ, Janin J, Baker D. (2011) Community-wide assessment of protein-interface modeling suggests improvements to design methodology. J Mol Biol 414: 289–302.
10.1016/j.jmb.2011.09.031
CAS PubMed Web of Science® Google Scholar
42Lee B, Richards FM (1971) The interpretation of protein structures: estimation of static accessibility. J Mol Biol 55: 379–400.
10.1016/0022-2836(71)90324-X
CAS PubMed Web of Science® Google Scholar
43Tsodikov OV, Record MT Jr., Sergeev YV (2002) Novel computer program for fast exact calculation of accessible and molecular surface areas and average surface curvature. J Comput Chem 23: 600–609.
10.1002/jcc.10061
CAS PubMed Web of Science® Google Scholar
44Chakrabarti P, Janin J (2002) Dissecting protein-protein recognition sites. Proteins 47: 334–343.
10.1002/prot.10085
CAS PubMed Web of Science® Google Scholar
45Hausman REC, Geoffrey M (2004) The cell: a molecular approach. Washington, DC: ASM Press.
Google Scholar
46McDonald IK, Thornton JM (1994) Satisfying hydrogen bonding potential in proteins. J Mol Biol 238: 777–793.
10.1006/jmbi.1994.1334
CAS PubMed Web of Science® Google Scholar
47Gupta PS, Mondal S, Mondal B, Islam RN, Banerjee S, Bandyopadhyay AK (2014) SBION: A program for analyses of salt-bridges from multiple structure files. Bioinformation 10: 164–166.
10.6026/97320630010164
PubMed Google Scholar
48Krissinel E, Henrick K (2007) Inference of macromolecular assemblies from crystalline state. J Mol Biol 372: 774–797.
10.1016/j.jmb.2007.05.022
CAS PubMed Web of Science® Google Scholar
49Liu S, Zhang C, Zhou H, Zhou Y (2004) A physical reference state unifies the structure-derived potential of mean force for protein folding and binding. Proteins 56: 93–101.
10.1002/prot.20019
CAS PubMed Web of Science® Google Scholar
50Zhou H, Zhou Y (2002) Distance-scaled, finite ideal-gas reference state improves structure-derived potentials of mean force for structure selection and stability prediction. Protein Sci 11: 2714–2726.
10.1110/ps.0217002
CAS PubMed Web of Science® Google Scholar
51Talley K, Ng C, Shoppell M, Kundrotas P, Alexov E (2008) On the electrostatic component of protein-protein binding free energy. PMC Biophys 1: 2.
10.1186/1757-5036-1-2
CAS PubMed Google Scholar
52Abagyan RA, Totrov MM, Kuznetsov DA (1994) ICM: A new method for protein modeling and design: applications to docking and structure prediction from the distorted native conformation. J Comp Chem 15: 488–506.
10.1002/jcc.540150503
CAS Web of Science® Google Scholar
53Abagyan RA, Totrov MM (2001) Rapid boundary element solvation electrostatics calculations in folding simulations: successful folding of a 23-residue peptide. Biopolymers 60: 124–133.
10.1002/1097-0282(2001)60:2<124::AID-BIP1008>3.0.CO;2-S
PubMed Web of Science® Google Scholar
54McDonald JH (2014) Handbook of biological statistics. Baltimore, MD: Sparky House Publishing.
Google Scholar
55 RStudio. RStudio: Integrated development environment for R v0.98.1091. Boston 2012.
Google Scholar
56Corder GW, Foreman DI (2014) Nonparametric statistics: a step-by-step approach. New York: Wiley.
Google Scholar

Citing Literature

Volume24, Issue9

September 2015

Pages 1486-1494

Linking structural features of protein complexes and biological function

Abstract

Introduction