Network analysis of protein dynamics

1. Introduction: topological networks of protein structures

• Networks represent complex systems as interacting nodes linked by weighted, and sometimes directed, edges.Edge weights can characterize interaction strength, affinity, intensity, or probability.Source: §1 · p. 1
• Many self-organized networks exhibit small-world structure, hubs, scale-free degree distributions, and overlapping, often hierarchical modules.Unlike random networks with Poissonian degree distributions and few hubs, scale-free networks have a non-negligible probability of high-degree hubs.Source: §1 · p. 1
• These network features are not universal: heterogeneity, sampling bias, or inadequate analysis can produce misleading patterns, so datasets and methods require scrutiny.Different modules may behave differently, and real-world networks often do not exhibit the generalized features in pure form.Source: §1 · p. 1
• Protein structure networks are also called amino-acid networks, residue-networks, or protein structure graphs, distinguishing them from protein-protein interaction networks.“Protein networks” is widely used for protein-protein interaction networks.Source: §1 · p. 2

2. Topological networks of protein structures

• Protein structure networks usually represent amino-acid side chains as nodes linked when their physical distance is below a cutoff, commonly 0.45–0.85 nm.Weighted links can reflect physical distance, but unweighted networks are more widely used; the approach was first used for structural comparison and similarity detection.Source: §2 · p. 2, Figure 1 · p. 3
• Protein structure networks have Poissonian rather than scale-free degree distributions, with fewer hubs attributed to the limited simultaneous binding capacity of amino-acid side chains.Each amino acid has a characteristic average degree; hydrophilic residues are strong hubs at high interaction cut-offs, whereas hydrophobic residues are weak hubs at low cut-offs, and hubs integrate secondary structures and increase thermodynamic stability.Source: §2 · pp. 2–3
• Central residues govern folding transitions and often occupy active or ligand-binding sites, while centrality redistributes between transitional and native structures and between active and inactive hemoglobin.Closeness, the inverse of mean shortest-path length, identifies many such residues; hydrogen-bond network plots also reveal central amino acids.Source: §2 · p. 3
• Assortativity and hierarchical hub structure occur in hydrophobic-amino-acid subnetworks, but not in hydrophilic or charged subnetworks, underscoring hydrophobic interactions in protein cores.Source: §2 · p. 3
• Protein structure networks are small-worlds across globular and fibrous proteins, and their small-worldness increases during folding as structures compact.Denser networks are associated with easier folding; these observations are mostly based on unweighted networks, and weighted-network assessments may differ.Source: §2 · p. 3
• Motifs are recurring assemblies of typically three to six amino-acid side chains, identifiable across evolutionary related protein sets of five or more proteins.Examples include catalytic triads and metal-coordination sites; a protein set may contain around 500 meaningful motifs, and ASSAM and DRESPAT support their search.Source: §2 · p. 4
• Protein structural modules correspond to domains, whose coordinated motion helps identify inter-domain residues important in functional regulation.Dense hydrogen-bond structures called stabilization centers have also been identified; domain-fold frequencies are scale-free, with popular stable, easily folded structures attributed to evolutionary selection.Source: §2 · p. 4

• Lacking conventional secondary structure helps intrinsically disordered proteins or regions bind and participate in recognition, while increasing the dynamics of proteins and complexes.Source: §3 · p. 4
• Protein disorder depends on timescale: chain flexibility causes fluctuations around an equilibrium conformation, unlike unstructured regions that lack an equilibrium conformation.Source: §3 · p. 4

4. Protein dynamics: quasi-harmonic movements, restricted relaxation and avalanches

• Protein-quakes after carbon monoxide photodissociation and other protein kinetics show scale-free timing and step sizes characteristic of restricted relaxation, which requires bond-breaking or local unfolding.By contrast, many motions after ligand binding require no bond rearrangements and are approximated as quasi-harmonic.Source: §4 · p. 4
• Hierarchical modular networks integrate correlated scale-free dynamics, complicating conformational rearrangements; phosphoglycerate kinase domain transitions are nonadditive, and distant-residue motions correlate in dihydrofolate reductase.Source: §4 · p. 4
• Side-chain fluctuations correlate with mean shortest-path length, with more central residues moving less; elastic-network models address omitted backbone constraints using αC atoms and harmonic contacts within 1 nm.Source: §4 · p. 5
• Conserved central residues are identified as allosteric signal-transmission elements, while concerted motions extending beyond enzyme active centers accompany catalysis.These findings come from elastic-network and conventional protein-structure-network analyses.Source: §4 · p. 5
• In spring networks, overconstrained regions correspond to rigid segments, whereas underconstrained regions correspond to flexible segments.Source: §4 · p. 5
• Hydrogen-bond perturbation propagation, represented in a simplified 2D HB-plot, revealed key dynamics features in cytochrome P450 and ligand-gated ion channels.Hydrogen-bond rearrangements also contribute to water's involvement in protein dynamics.Source: §4 · p. 5

5. Protein dynamics: Water as a lubricant

• Water’s fluctuating hydrogen bonds with peptide bonds and side chains can lower energy-landscape saddles, enabling otherwise forbidden conformational transitions.Water-induced fluctuations decrease as folding proceeds.Source: §5 · p. 5
• A surface water monolayer and percolating hydrogen-bond network restore biomolecular dynamics, yet residual enzyme activity can persist at very low hydration.Source: §5 · pp. 5–6
• Some protein motions are water-slaved, while others are solvent-independent; most enzymatic function stops without water, although pre-stabilized dry proteins can retain activity as conformational changes freeze.Analyzing hydrogen-bond networks across hydration levels is identified as future work.Source: §5 · p. 6

6. Energy and conformational networks in the description of protein dynamics

• Energy networks represent local energy minima as nodes and transition-state links, and show small-world and scale-free character.A weighted energy network used modules to identify landscape basins; weighted small-world assessment remains future work. The large number of minima grows exponentially with residues involved, motivating simpler renormalized representations.Source: §6 · p. 6
• Scale-free degree distributions occurred in enthalpy-dominated energy-network modules, while entropy-dominated modules showed Gaussian degree distributions.Energy-network complexity, roughly the number of energy basins, has been proposed to distinguish easy folders from proteins prone to becoming trapped in conformational possibilities; defining its numerical link to network complexity remains difficult.Source: §6 · p. 6
• Treating all possible conformations as local minima and identifying basins as primary modules offers a way to represent shallow, fluctuating energy basins in hierarchical networks.Local minima need not be sharp features; shallow basins can contain numerous fluctuating minima.Source: §6 · p. 7
• Energy-network links may be directed, with greater energy differences between neighboring minima corresponding to more directed links.Source: §6 · p. 7
• Energy-network small-worldness may help explain protein dynamism, while hierarchical traps account for stretched early-folding kinetics and aging at cryogenic temperatures.The hierarchical energy landscape gives the network a fractal-like structure resembling Apollonian networks.Source: §6 · p. 7
• Conformational networks connect protein conformations through transitions and, like energy networks, have been used to predict native structures and assess transition probabilities.The two representations describe essentially the same protein-state ensemble using different data and rules.Source: §6 · p. 7
• Analyzing the dynamics of protein structural networks across conformations is identified as an important future development.Source: §6 · p. 7

Summary and perspectives

• Protein structure and conformational networks are small worlds, while central residues and hubs help integrate structure and govern protein stability, motion, and signaling.Hydrophobic amino acids appear especially important to structure-network integration. Central residues often lie in active or ligand-binding sites, which can trigger protein-wide motion during catalysis or signal transduction.Source: §7 · p. 7
• Modular network analysis identifies inter-modular residues that can govern domain-boundary transitions and can simplify rugged energy landscapes.Renormalizing conformational or energy networks may help distinguish easy folders from proteins with many folding traps, making the landscapes more computationally and cognitively tractable.Source: §7 · pp. 7–8
• Future work should compare link weights and network-building rules systematically rather than relying on cutoff distances and unweighted links.The review suggests that reanalyzing small-worldness in weighted networks may yield new findings.Source: §7 · p. 8
• Refined analysis of hierarchical and overlapping modules could reveal functionally important residues, while modular analysis could support graded, multilevel centrality measures.The review proposes centrality indices that account for weights and all topological levels to identify modular centers, inter-modular bridges, and elements of overlapping regions; hierarchical and overlapping module analysis may help identify residues governing enzyme activity, allostery, motor function, signal transduction, and stability.Source: §7 · p. 8
• Weighted and directed links combined with hierarchical modular analysis may make rugged conformational and energy landscapes more comprehensible.Source: §7 · p. 8
• Hypergraphs representing non-paired interactions may enable more refined analyses of protein structure and transitions.Source: §7 · p. 8
• Analyzing the dynamics and evolution of protein structural networks could reveal correlated network regions corresponding to correlated protein motions.The review also proposes ‘protein games’; perturbational analysis of cooperative conformational-network regions has provided evidence for central arrangement of active centers.Source: §7 · p. 8
• The review suggests protein network studies are near a ‘tipping point’ of increased applicability for new concepts.The authors hope their review contributes to the field’s expansion.Source: §7 · p. 8