Protein Structure: Biophysical Levels of Folding, Force Fields, and Conformational Stability

Introduction

Proteins are the molecular machines of life, and their three-dimensional structures dictate their biological functions. The process by which a linear polypeptide chain adopts a unique, functional conformation is known as protein folding. This process is governed by a complex interplay of physical forces, including van der Waals interactions, hydrogen bonding, electrostatic forces, and the hydrophobic effect [1, 2]. Understanding the biophysical basis of protein folding is essential for interpreting structure-function relationships, predicting the effects of mutations, and designing therapeutic interventions in veterinary medicine and diagnostics [3, 4]. This article provides an exhaustive review of the hierarchical levels of protein folding, the force fields used to model them, and the thermodynamic and kinetic principles that determine conformational stability.

Hierarchical Levels of Protein Folding

Protein structure is traditionally described at four levels: primary, secondary, tertiary, and quaternary. However, modern biophysics also recognizes the importance of intrinsically disordered regions and phase-separated states [5, 6].

Primary Structure

The primary structure is the linear sequence of amino acids linked by peptide bonds. This sequence encodes all the information necessary for folding [7, 8]. Mutations in the primary sequence can alter folding pathways and stability, as demonstrated in studies of viral envelope proteins where single amino acid substitutions affect receptor binding and immune evasion [9, 10].

Secondary Structure

Local hydrogen bonding patterns give rise to regular secondary structures: alpha-helices, beta-sheets, and turns. These elements are stabilized by backbone amide-carbonyl interactions [2, 11]. Secondary structure prediction algorithms, such as PSIPRED, use neural networks to assign these elements from sequence alone [12]. The accuracy of such predictions has improved dramatically with deep learning [13].

Tertiary Structure

Tertiary structure refers to the global three-dimensional arrangement of a single polypeptide chain. It is stabilized by side-chain interactions, including hydrophobic packing, salt bridges, disulfide bonds, and hydrogen bonds [1, 14]. The folding process involves a cooperative transition from an unfolded ensemble to a compact native state, often described by a funnel-shaped energy landscape [4, 15]. Experimental techniques such as cryo-electron microscopy (cryo-EM) and X-ray crystallography provide atomic-resolution structures [16, 17]. Computational methods like homology modeling (e.g., MODELLER, SWISS-MODEL) and deep learning (e.g., AlphaFold) have revolutionized tertiary structure prediction [7, 18, 19].

Quaternary Structure

Many proteins function as oligomeric complexes. Quaternary structure describes the arrangement of multiple subunits. The assembly is driven by complementary surfaces and often involves allosteric regulation [14, 20]. For example, the RACK1 scaffolding protein adopts a seven-bladed beta-propeller that mediates protein-protein interactions in signaling pathways [20]. Bacterial microcompartment shell proteins self-assemble into polyhedral shells through charged residue interactions [14].

Intrinsically Disordered Regions and Phase Separation

Not all proteins fold into stable globular structures. Intrinsically disordered proteins (IDPs) lack persistent secondary or tertiary structure under physiological conditions [6, 5]. They often undergo folding-upon-binding when interacting with partners [6]. Liquid-liquid phase separation, driven by multivalent interactions and RNA G-quadruplexes, is a key mechanism for organizing cellular compartments [5]. The conformational dynamics of IDPs can be modulated by small molecules, as shown for the SARS-CoV-2 nucleocapsid protein [21].

Force Fields in Molecular Modeling

Molecular dynamics (MD) simulations rely on force fields to compute the potential energy of a protein system as a function of atomic coordinates [11, 22]. Force fields are parameterized sets of equations that describe bonded interactions (bonds, angles, dihedrals) and non-bonded interactions (van der Waals and electrostatic) [2, 23].

Classical Force Fields

Classical force fields such as AMBER, CHARMM, and OPLS use fixed partial charges and harmonic potentials. They are widely used for simulating protein dynamics and folding [11, 22]. The accuracy of these force fields depends on the parameterization of dihedral angle potentials and the treatment of solvation [2]. Recent developments include polarizable force fields that account for electronic polarization [23].

Coarse-Grained Force Fields

For large systems or long timescales, coarse-grained models reduce computational cost by grouping atoms into beads. The MARTINI force field is a popular example. Coarse-grained simulations have been applied to study protein self-assembly and membrane interactions [14, 24].

Quantum Mechanical and Hybrid Methods

Quantum mechanical (QM) methods provide accurate descriptions of bond breaking and formation, but are computationally expensive. Hybrid QM/MM approaches combine a QM region for the active site with a MM region for the rest of the protein [1, 22]. These methods are essential for studying enzyme catalysis and metal-binding proteins [1].

Machine Learning Potentials

Recent advances have introduced machine learning potentials trained on quantum chemical data. These potentials achieve near-QM accuracy at MM cost [23, 13]. They are particularly useful for modeling conformational transitions and ligand binding [9].

Conformational Stability

Conformational stability is the thermodynamic preference of a protein for its native state over unfolded or misfolded states. It is quantified by the Gibbs free energy of folding (ΔG_fold) [4, 15].

Thermodynamic Principles

The folding equilibrium is governed by the balance between enthalpy (favorable interactions) and entropy (chain conformational entropy and solvent entropy). The hydrophobic effect, driven by the release of water molecules from nonpolar surfaces, is a major contributor to folding stability [2, 4]. The stability of a protein can be perturbed by temperature, pH, and denaturants [2].

Kinetic Aspects

Folding kinetics describe the rate at which a protein reaches its native state. Many proteins fold via a two-state mechanism, while larger proteins may populate intermediate states [6, 15]. Chaperones assist folding by preventing aggregation [2]. The folding landscape can be rugged, with local minima corresponding to misfolded states [4].

Experimental Characterization

Biophysical techniques such as circular dichroism (CD), fluorescence spectroscopy, and differential scanning calorimetry (DSC) are used to monitor folding transitions [1, 2]. For example, the unfolding of bovine serum albumin by Nb₂C nanochaperones was studied using spectroscopic methods [2]. Surface plasmon resonance (SPR) and isothermal titration calorimetry (ITC) measure binding affinities and conformational changes [21].

Computational Prediction of Stability

Computational methods can predict the effect of mutations on protein stability. Tools like FoldX, Rosetta, and machine learning models use energy functions to estimate ΔΔG [3, 4]. The biophysical fitness landscape approach designs sequences that trap viral evolution by destabilizing essential proteins [4]. Structural feature-based machine learning has been applied to predict protein-protein interface stability [3].

Computational Structure Prediction

The field of protein structure prediction has undergone a revolution with the advent of deep learning. AlphaFold2 achieved atomic accuracy in the CASP14 experiment, solving the classical protein folding problem for single domains [7, 25]. The AlphaFold Protein Structure Database now contains over 214 million predicted structures, covering most known protein sequences [26, 27]. Language models such as ESMFold enable rapid structure prediction without multiple sequence alignments, achieving up to 60x speedup [8, 10]. Other methods like I-TASSER and Phyre use threading and ab initio modeling [28, 29]. Structure comparison tools like DALI, TM-align, and Foldseek allow searching of structural databases [30, 31, 32]. De novo design methods such as RFdiffusion generate novel protein backbones for therapeutic applications [33].

The following Mermaid diagram summarizes the hierarchical folding levels and the computational methods used at each stage.

flowchart TD
    A[Primary Sequence], > B[Secondary Structure Prediction]
    B, > C[Tertiary Structure Prediction]
    C, > D[Quaternary Structure Assembly]
    D, > E[Functional Dynamics]
    
    B, > F[PSIPRED, Deep Learning]
    C, > G[AlphaFold, I-TASSER, MODELLER]
    D, > H[Rosetta, RFdiffusion]
    E, > I[MD Simulations, Normal Mode Analysis]
    
    F, > J[Force Fields: Classical, QM/MM, ML]
    G, > J
    H, > J
    I, > J
    
    J, > K[Conformational Stability Analysis]
    K, > L[Experimental Validation: CD, DSC, Cryo-EM]

Applications in Veterinary Medicine and Diagnostics

Understanding protein structure and stability is critical for veterinary virology and diagnostics. For example, predicting the structure of viral glycoproteins informs vaccine design and antibody neutralization [9, 33]. The conformational dynamics of viral envelope proteins affect host range and immune evasion [9]. Computational modeling of host-pathogen protein-protein interactions can identify targets for antiviral therapy [3, 4]. In diagnostics, structure-based design of capture antibodies and antigens improves assay sensitivity [21]. The ability to predict the impact of mutations on protein stability is essential for monitoring emerging viral variants in animal populations [4, 13].

Conclusion

Protein folding is a hierarchical process governed by fundamental physical forces. Modern computational methods, from classical force fields to deep learning, have enabled accurate prediction of protein structures and stability. These tools are transforming veterinary structural biology, allowing researchers to understand disease mechanisms and develop novel interventions. Continued integration of biophysical principles with machine learning will further advance the field.

References

[1] Ben-David A, Onisuru O, Oyiogu B, et al. Heterologous Expression, Purification, and Biophysical Characterisation of the Cobalt-Dependent Nitrile Hydratase from Rhodococcus rhodochrous ATCC BAA-870. Biochimie. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42364840/

[2] Li N, Ye Y, Zhou T, et al. Inhibitory effect and mechanism of Nb(2)C nanochaperone on bovine serum albumin unfolding. Spectrochim Acta A Mol Biomol Spectrosc. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42364481/

[3] Topuz T, Erdem Z, Bisgin H, et al. Structural feature-based machine learning benchmarking for protein interface prediction. Sci Rep. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42365029/

[4] Mohanty V, Shakhnovich EI. Biophysical fitness landscape design traps viral evolution. Proc Natl Acad Sci U S A. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42361039/

[5] Carey JL, Hayashi M, Welebob E, et al. RNA G-quadruplexes function as a tunable switch of FUS phase separation. Nucleic Acids Res. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42363764/

[6] Kjaer LF, Ielasi FS, Winbolt T, et al. Alternative Splicing of a Structured Partner Alters the Folding-Upon-Binding Trajectory of an Intrinsically Disordered Protein. J Am Chem Soc. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42361232/

[7] Jumper J, Evans R, Pritzel A, et al. Highly accurate protein structure prediction with AlphaFold. Nature. 2021. URL: https://www.semanticscholar.org/paper/dc32a984b651256a8ec282be52310e6bd33d9815

[8] Lin Z, Akin H, Rao R, et al. Evolutionary-scale prediction of atomic level protein structure with a language model. bioRxiv. 2022. URL: https://www.semanticscholar.org/paper/c49a0912595a1cc70aab63524f64ed08c92194a8

[9] Cao Y, Im W. Conformational variability of HIV-1 Env trimer and viral vulnerability. Elife. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42360802/

[10] Chowdhury R, Bouatta N, Biswas S, et al. Single-sequence protein structure prediction using language models and deep learning. Nature Biotechnology. 2022. URL: https://www.semanticscholar.org/paper/7698498dcb14db063154f4c955fc041114d1960d

[11] Zhang K, Zhang J, Chen N, et al. Dynamic conformations of fluorophores on self-labeling protein tags. Biophys J. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42363595/

[12] McGuffin L, Bryson K, Jones DT. The PSIPRED protein structure prediction server. Bioinform. 2000. URL: https://www.semanticscholar.org/paper/b36b9a86a72289ff7ba9e0c77a0fef115560bb60 *** Disclaimer: This article is for educational and informational purposes only. It is not intended to substitute for professional veterinary advice, diagnosis, treatment, or regulatory guidance. Always consult a licensed veterinarian or qualified specialist regarding animal health, disease diagnosis, and therapeutic decisions.

[13] Bertoline L, Lima AN, Krieger J, et al. Before and after AlphaFold2: An overview of protein structure prediction. Frontiers in Bioinformatics. 2023. URL: https://www.semanticscholar.org/paper/797211d61253b571d29c08c175e7904156cc321c

[14] Gomez A, Mehrafrooz B, Waltmann C, et al. Engineering the Self-Assembly of Bacterial Microcompartment Shell Proteins via Charged Mutations. ACS Nano. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42363890/

[15] Wu KE, Yang KK, van den Berg R, et al. Protein structure generation via folding diffusion. Nature Communications. 2022. URL: https://www.semanticscholar.org/paper/15586fec836ffcb60eba81491e04c225e6914aac

[16] Karki S, Lu X, Paatero AO, et al. Cryo-EM structure of the Arabidopsisthaliana ribosome in translating and non-translating states. Structure. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42361795/

[17] Broutzakis G, Pyrris Y, Akrani I, et al. Cryo-EM of the eukaryotic purine transporter UapA demonstrates intramolecular and lipid regulation of transport. Proc Natl Acad Sci U S A. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42361038/

[18] Webb BM, Sali A. Comparative Protein Structure Modeling Using MODELLER. Current Protocols in Bioinformatics. 2016. URL: https://www.semanticscholar.org/paper/36abaca9db1327fdff3dfdd07a58aae61eede7d2

[19] Arnold K, Bordoli L, Kopp J, et al. The SWISS-MODEL workspace: a web-based environment for protein structure homology modelling. Bioinform. 2006. URL: https://www.semanticscholar.org/paper/c21252cac5b6c33012d2a071433f79ab0beeb0fb

[20] Adams DR, Ron D, Kiely PA. RACK1, A multifaceted scaffolding protein: Structure and function. Cell Communication and Signaling. 2011. URL: https://www.semanticscholar.org/paper/4c4c71a43d2eac9e25077de578e63be1866829d1

[21] Vermaak TL, Mosebi S, Gildenhuys S. Biophysical analysis reveals low‑affinity, multivalent interactions and conformational modulation of the SARS‑CoV‑2 nucleocapsid protein by berberine and palmatine. J Biomol Struct Dyn. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42364124/

[22] Mohtashim SI, Sajjan M, Kais S. Continuous-Time Quantum-Walk Centrality for Protein Residue Interaction Networks. J Am Chem Soc. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42361045/

[23] Senior A, Evans R, Jumper J, et al. Improved protein structure prediction using potentials from deep learning. Nature. 2020. URL: https://www.semanticscholar.org/paper/3a083d843f891b3574494c385699c21766ce8b7a

[24] Chen X, Chen J, Zhang J, et al. Cross-Scale Biomimicry of Resilin and Resilin-Like Proteins: From Phase Separation and Dityrosine Photocrosslinking to Tissue Repair and Soft Robotic Devices. Acta Biomater. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42361867/

[25] Kryshtafovych A, Schwede T, Topf M, et al. Critical Assessment of Methods of Protein Structure Prediction (CASP) – Round XIV. Proteins: Structure, Function, and Bioinformatics. 2021. URL: https://www.semanticscholar.org/paper/11c0c566140dd607c1c4c8f86f14e3dd6c238c1f

[26] Váradi M, Anyango S, Deshpande M, et al. AlphaFold Protein Structure Database: massively expanding the structural coverage of protein-sequence space with high-accuracy models. Nucleic Acids Res. 2021. URL: https://www.semanticscholar.org/paper/86c048c4d91066610beba9099b7d23bdae29caab

[27] Váradi M, Bertoni D, Magana P, et al. AlphaFold Protein Structure Database in 2024: providing structure coverage for over 214 million protein sequences. Nucleic Acids Research. 2023. URL: https://www.semanticscholar.org/paper/a938303f20b7fcde2bfdb7ca95ff5b69739430d3

[28] Yang J, Yan R, Roy A, et al. The I-TASSER Suite: protein structure and function prediction. Nature Methods. 2014. URL: https://www.semanticscholar.org/paper/636b17db1066f5c5b028cec35bc3477a051d10f8

[29] Kelley L, Sternberg M. Protein structure prediction on the Web: a case study using the Phyre server. Nature Protocols. 2009. URL: https://www.semanticscholar.org/paper/97facaf1d4cc3647afb95c91e3984963ee153137

[30] van Kempen M, Kim SS, Tumescheit C, et al. Fast and accurate protein structure search with Foldseek. bioRxiv. 2022. URL: https://www.semanticscholar.org/paper/655fa18c26267f33084647b30fb0964ec199cd97

[31] Holm L, Sander C. Protein structure comparison by alignment of distance matrices. Journal of Molecular Biology. 1993. URL: https://www.semanticscholar.org/paper/9c21b7300178db8b18e2e289db810284f1575c3b

[32] Zhang Y, Skolnick J. TM-align: a protein structure alignment algorithm based on the TM-score. Nucleic Acids Research. 2005. URL: https://www.semanticscholar.org/paper/08a05b9e8548354e24542c2b5f68b1432b50b27d

[33] Watson JL, Juergens D, Bennett N, et al. De novo design of protein structure and function with RFdiffusion. Nature. 2023. URL: https://www.semanticscholar.org/paper/eb8c0df75993c1c19c51cde9345e45fc260f661c

[34] Tunyasuvunakool K, Adler J, Wu Z, et al. Highly accurate protein structure prediction for the human proteome. Nature. 2021. URL: https://www.semanticscholar.org/paper/1fb3ad9969245795f268636eff9a145337144718

[35] Krissinel E, Henrick K. Secondary-structure matching (SSM), a new tool for fast protein structure alignment in three dimensions. Journal. URL: https://www.semanticscholar.org/paper/a9540be379f836740c19238594d500b73fd18cf8