Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2023 Oct 19.
Published in final edited form as: Proc (Int Conf Comput Sci Comput Intell). 2023 Aug 25;2022:572–577. doi: 10.1109/csci58124.2022.00108

Collagen a1(XI) structure prediction by Alphafold 2

Abu Sayeed Chowdhury 1, Julia Thom Oxford 2,*
PMCID: PMC10586751  NIHMSID: NIHMS1908254  PMID: 37860747

Abstract

Collagen α1(XI) is a minor fibrillar collagen involved in the critical regulation of collagen fibrils such as nucleation, assembly, and regulation of fibril diameter. The amino propeptide domain of the collagen α1(XI) is retained on the surface of the collagen fibril for an extended period of time and may play a crucial role in the interaction with extracellular matrix glycosaminoglycans and other proteins during the process of fibrillogenesis. Understanding the mechanism of action of this protein will ultimately help us understand the organization and assembly of the extracellular matrix that underlies the structural integrity of connective tissues.

Keywords: collagen α1(XI), amino propeptide domain, Alphafold 2, fibrillogenesis, glycosaminoglycans

I. Introduction

The main constituents of connective tissue in mammals are collagens, a family of fibrous proteins. Tropocollagen is the basic structural unit of collagen comprised of 3 intertwined left-handed helices that ultimately form a right-handed triple helix [1][2]. A repetition of the triplet (Gly-Xaa-Yaa) is observed where approximately 30% of the residues are either prolines or hydroxyprolines [3]. Collagen type XI is minor fibrillar collagen that is detected on the surface of heterotypic collagen fibrils composed of type II, type IX, and type XI. Type XI collagen plays a crucial role in the regulation of nucleation, and assembly, and in determining the final diameter of collagen fibrils [4][5]. Deficiencies in collagen type XI expression cause chondrodystrophy in the Cho/Cho mouse and Sickler syndrome type 2 in humans [5]. Collagen XI N-terminal pro-peptide domain (Npp) forms a globular “head” while the variable region forms a linker region between the Npp and the minor helix [6] (fig. 1). To date, there are approximately 28 distinct types of collagens, and among these, sixteen share a common non-collagenous domain known as a thrombospondin-like domain (Thsp1 or TSPN domain) [7–10]. Similarity exists between the TSPN domain and the LNS domain (laminin A, neurexin, and sex hormone binding globulin [11]. In vitro studies suggest that the regulatory behavior of collagen XI is dependent on the complex formation within the amino-terminal domain (NTD) of the alpha 1 chain, which can be proteolytically removed, albeit is slow when compared to the NTD of the other alpha chains [5][12–13]. Thus, its existence on the surface of collagen fibrils may impose steric hindrance and limit further deposition of collagen triple helices to the growing fibril. This may result in lateral growth restriction, however, at the same time, it may provide stabilizing interactions between collagen fibrils and/or other extracellular matrix components [4].

Fig. 1:

Fig. 1:

Open in a new tab

Schematic representation of the Npp α1 (XI), variable region (p6a, p6b, p8), and adjacent minor triple helix. The dimensions were determined by previously by transmission electron microscopy [6]. The α2 (XI) and α3 (XI) chains are shown. The diagram is adopted from [8].

Currently, there is no experimentally determined protein structural information for the Npp domain of collagen α1(XI), however, the NC4 domain structure of collagen IX has been experimentally determined, and these domains are homologous [10]. This domain is a member of the laminin, neurexin, and sex hormone binding globulin (LNS) domain [14]. LNS domains may interact with heparin and heparan sulfate, which is present in several proteoglycans of the extracellular matrix. The NC4 domain of collagen type IX interacts with heparan sulfate, and this domain is also exposed on the collagen surface [15]. It has been hypothesized that heparan sulfate or similar glycosaminoglycans may fulfill the bridging function between adjacent collagen fibrils and the cell surface, which may explain the role of the LNS domain containing collagens in the process of fibrillogenesis. Perhaps it may also explain the apparent role of this LNS domain in matrix-matrix interactions as well as the interactions between cells and their immediate environment. If these are true, then the LNS domain could play a central role in tissue integrity [11].

Npp of the α1 (XI) chain is comprised of 223-amino-acids, and this sequence was used for a BLASTP search on PubMed in this study [14]. The search returned the crystal structure of the N-terminal NC4 domain of collagen IX (PDB ID: 2UUR), which is a member of the LNS domain family [10]. This was the best result based on the E value, percent identity, sequence similarity, query coverage, and max score. In the past, we considered the template from the crystal structure of 2UUR and laminin α2 chain (PDB ID: 1dyk) to build a homology model using Modeller software [16–18]. In this current study, we report the structure of the Npp α1 (XI) collagen built with a neural network-based model, Alphafold 2, which may predict protein structures more accurately than previously possible [19]. To validate the best structure based on stability and typical predicted behavior in an aqueous environment among three candidate structures, a molecular dynamics simulation over a period of 60 nanoseconds (ns) was performed using the GROMACS molecular dynamics package [20].

II. Materials and methods

A. Building 1dyk template-based homology structure with Modeller

The LG 5 domain (residues 2934- 3117) was selected from the crystal structure of α2 laminin (PDB: 1dyk). According to previous work [11], good sequence alignment was found for 183 amino acids (TEEA to PHIDE). The protein structure was processed with ChimeraX [21] to remove bound water molecules and calciumions. This portion of the protein was used to build the homology model using Modeller (version 10.3). The rest of the 40 amino acids were subjected to ab initio modeling using the Rosetta server [22]. Subsequently, the two structures were joined together with ChimeraX. Loop refinement was performed with the Modeller software plug-in in Chimera. Residues in the outlier region in the Ramachandran plot were detected and solved via energy minimization with Chiron [23], and residue clashes were amended via Chimera.

B. Building 2UUR template-based homology structure with Modeller

The x-ray crystallographic structure of 2UUR was downloaded from the RCSB PDB and the structure was processed in Chimera to remove crystallized SO4, Zn, and 1,2-ethanediol. The remaining structure was used in Modeller (version, 10.3) for homology model building. The structure derived from Modeller was then subjected to evaluation for the areas of high energy, restricted areas in the Ramachandran plot, and residue clashes. Loop refinement was done using the Modeller plug-in in Chimera and energy minimization was done using Chiron [23] as well as in Chimera.

C. Building Npp structure using Alphafold 2

The research computing facility of Boise State University was used for the operation of Alphafold 2. A monomeric function was used to predict the structure. Among the structures, the highest model confidence based on the predicted local distance difference test (pLDDT) score was selected and used for further work. The Alphafold 2-derived meta-data was analyzed for visualization of specific areas of the protein to determine the model confidence level. Two previously described disulfide bonds between residue Cys25 – Cys 207 and Cys 146 -Cys 200 of the protein were supported by the structural prediction and these were introduced manually in Chimera for the Alphafold-derived structure [10–11, 14, 24]. MD simulation data were analyzed and plotted using Microsoft Excel and GraphPadPrism.

D. Molecular dynamics simulation

The protein structures were processed for protonation state and adjusted as per the solvent system and histidine (His) residues were considered to be in a neutral state. The 223 amino acid residues yield an overall negative charge, so 4 sodium ions were added to neutralize it. A periodic cubic simulation box was created in which the protein was enclosed. The box was filled with water (spc216) as the solvent. OPLS-AA/L all-atom force field [25] was used in this experiment. Next, the energy minimization was performed using a maximum force value of 10 kJ/nm/mol using the steepest descent method for 60,000 steps. The equilibration of the system was performed in two steps. First, we used the “isothermal-isochoric” or “canonical ensemble”, followed by the “isothermal-isobaric” ensemble. Both of the equilibration steps were run for 100 ps with temperature and pressure set at 300 K and 1 atm, respectively. For the particle mesh Ewald (PME), long-range electrostatic interactions were calculated. When the desired temperature and pressure were well-equilibrated, the position restraints were released and the MD run was initiated.

III. Results and Discussion

For modeling purposes, 223-residue amino propeptide was used, which is a collagen XI α1 chain extending from the signal peptide cleavage site to the amino acid immediately preceding the variable region (fig. 1.).

After building the three structures they were superimposed using the molecular visualization software Chimera and the match-align function was used to compare the structural features including alpha-helices, beta-sheet, and loops (fig. 2). From fig. 2, it is apparent that the Npp structure generated by the Modeller based on the 2UUR template and the structure from Alphafold 2 are similar. Alphafold 2 utilized 20 template matches, among them the artificial intelligence-based system used the best four templates. The four templates used by Alphafold 2 were the N-terminal NC4 domain of collagen IX (PDB ID: 2UUR), Thrombosopondin-1 N-terminal domain in complex with Arixtra (PDB ID: 1ZA4) [14], thrombospondin-1 N-terminal domain complexed with synthetic pentameric heparin (PDB ID: 2ERF) [26], and the N-terminal domain of sex hormone binding globulin (PDB ID: 1KDK) [27]. The N-terminal NC4 domain of collagen IX is similar to the LNS domain and it conserves the structural features and functionality of the TSPN-1 domain [10]. Interestingly, the best three templates were structural proteins interacting with heparin or heparan sulfate [10, 26]. Disulfide bonds between residue Cys25 and Cys207 and between Cys146 and Cys200 were introduced using Chimera (Fig.3). Moreover, several salt bridges and attractive charge interactions have been identified between amino acids lys 127 and asp 125, arg 130 and asp 206, arg 39 and asp 183, arg 28 and asp 35, arg 156 and asp 62, lys 138 and glu 158, and arg 173 and glu 176.

Fig. 2:

Fig. 2:

Open in a new tab

Alignment of the three tertiary structures based on the structural features. Green indicates β-strands, yellow indicates α-helix and the rest indicates loops. The gaps are indicated by (.).

Fig. 3:

Fig. 3:

Open in a new tab

(a). Alphafold 2 confidence level scores (pLDDT), (b) & (c). Color-coded structure of the collagen 11 α1-Npp and 180° rotated structure, (d). a disulfide bond between residue Cys- 25 and Cys- 207, (e). a disulfide bond between Cys-146 and Cys- 200 which were introduced manually by Chimera.

The structure from Alphafold 2 showed the highest accuracy compared to the other two structures using a Ramachandran plot (0% questionable observation, approximately 98% highly preferred observation, and 2% preferred observation). The pLDDT scores of the Alphafold 2-derived structure are presented in Table 1.

TABLE I.

Secondary Structural Features of Alphafold 2 Derived Structure

Per-residue confidence score (pLDDT) Residues, number, and location Pct. (%)
Very low (pLDDT < 50) Amino acid residue 213 to 223: loop 4.9
Low (70 > pLDDT >50) Lys 32, Thr 210, Ser 211: loop 1.4
Confident (90 > pLDDT > 70) Ala 1, Ser 27, Arg 28, Gly 45, Pro 55, Gly 56, Gly 57, Ile 174, Leu 175, Glu 176, Thr 177: loop
Asp 198, Asp 201, Tyr 202, Pro 112: alpha-helix 		7.2
Very high (pLDDT > 90) Rest of the residues 86.6
Open in a new tab

Alphafold 2 predicted 4 of the amino acid residues on the alpha-helix with a pLDDT score between 90 to 70. It has been found that Alphafold 2 slightly over-predicts the alpha-helices and beta-sheets in proteins based on DSSP (Defined Secondary Structure Protein) analysis. Generally, Alphafold 2 is considered to be a good predictor of the structure of loops, more precisely loops with less than 10 residues in length. But with the increasing length, the accuracy of the prediction drops based on the RMSD, TM-score, and DSSP [28]. However, previous circular dichroism spectral analysis indicated that the α1-Npp contains 33% beta-sheet [6], accurately matching our predicted structure of the protein (32.7%). The β-sheets were within a very high confidence level. The model predicted a total of 12 antiparallel β-beta-strands, 3 small α-helices, and 3 310 α-helices. The β-sheets formed a sandwich. It has been well documented that the 310 α-helices are dynamic and intrinsically less stable compared to the α-helices, and are rare [29].

RMSD values of the protein backbone were used to indicate the conformational stability of the protein structures. Among the three structures, the rmsd found from Alphafold 2 is the lowest, indicating that it is the most stable structure (fig. 4) [30]. The specific structure had a mean rmsd of 0.23 nm which is lower than the 1dyk template-based Modeller structure (0.69 nm) and 2uur template-based Modeller structure (0.35 nm).

Fig 4:

Fig 4:

Open in a new tab

Time evolution of rmsd of the three structures.

Next, the residue-specific fluctuation of a protein structure indicated by the individual residue variance across the sequence may indicate the flexibility of a specific region of the protein [10]. From the Alphafold 2-derived structure, it was apparent that the flexible regions are located in loops (fig 5). Residue-specific flexibility was observed from residues 25 to 35 (fig 5). Flexibility was observed in another loop region (residue 55 to residue 59). The most flexible residues are 117-QEPHIDE-223 for which the pLDDT score is very low to low. However, the bone morphogenic protein 1 (BMP-1) processing site consensus is between QA and QE within this region. The rmsf result agreed with the structural model as the pLDDT score of Alphafold 2 is an indicator of flexibility, and disordered regions [19]. The mean rmsf was the lowest for the Alphafold 2-derived structure among the three structures and precluding the last 7 residues 117-QEPHIDE-223 of the Alphafold 2-derived structure, the standard deviation was also the lowest. The presence of polar amino acids at the surface, and non-polar amino acids at the core may support a higher number of internal hydrogen bonds and better stability (fig.6) [31].

Fig. 5:

Fig. 5:

Open in a new tab

Time evolution of rmsf of the three structures.

Fig. 6:

Fig. 6:

Open in a new tab

Time evolution of total intra-protein hydrogen bonds of the three structures.

Most frequently, loops of a protein are found near the surface of the protein which exposes them to solvents and other proteins [32]. This feature enables loops to play crucial roles in protein structure such as shielding a hydrophobic core or binding to other proteins or ligands [33–37]. Among the three structures, the Alphafold 2-derived structure had the lowest solvent-accessible surface area (SASA) (fig. 7), which may be due to the flexible loops shielding the hydrophobic core. Further, experimentation is required to investigate this hypothesis. However, there is an opposite trend observed between the increasing number of hydrogen bonds and decreasing SASA and vice-versa (fig. 9).

Fig. 7:

Fig. 7:

Open in a new tab

Time evolution of SASA of the three structures.

Fig. 9:

Fig. 9:

Open in a new tab

The effect of an increasing number of average hydrogen bonds with the change of SASA.

The radius of gyration (Rg) was used to indicate the conformational transition and the compactness of a protein [30]. Among the three structures, the Alphafold 2-derived structure had the lowest Rg (fig. 8). Moreover, there was an opposite trend between the total number of intra-protein hydrogen bonds and the average Rg (fig. 10).

Fig 8:

Fig 8:

Open in a new tab

Time evolution of Rg of the three structures.

Fig. 10:

Fig. 10:

Open in a new tab

The effect of an increasing number of average hydrogen bonds with the change of radius of gyration.

The structural stability of the protein depends on the interactions among amino acids which support the formation of a secondary structure [30]. Comparing the three structures, the Alphafold 2-derived structure had the highest number of hydrogen bonds (fig. 6 & 11).

Fig 11:

Fig 11:

Open in a new tab

Intra-protein hydrogen bonds shown in purple dashed lines.

The Alphafold 2-derived structure showed that most of the intra-protein hydrogen bonds are formed within the β-strands and a smaller number of hydrogen bonds within the loops. A total of 182 intra-protein hydrogen bonds were found, and among them, 152 were strong hydrogen bonds. The hydrogen bonds on the loops were longer compared to the hydrogen bonds within the β-sheets. The shorter, stronger hydrogen bonds [31] make the core more compact, agreeing with the molecular dynamics simulation results. On the contrary, the hydrogen bonds on the loops and α-helices were longer and weaker compared to the shorter hydrogen bonds [38]. The compact core contained the majority of hydrophobic amino acids compared to the loops as per the analysis by ProtParam [39].

IV. CONCLUSION

In this work, we built three structures of collagen α1 (XI) Npp domain, based on the 1dyk template and 2uur template. Two of them were homology-based models built by Modeller software. The third structure was built by the neural network-based model Alphafold-2. Comparative analysis was done using MD simulation and structural feature analysis. In total, 6.3% of the total amino acid residues were in the very low to low confidence level prediction by Alphafold 2. These residues were located at the terminus of the collagen α-1 (XI) Npp immediately adjacent to the variable region. The predicted β-sheet content matched the experimental data. MD simulation also indicated that the Alphafold 2 structure has the lowest root mean square deviation, root mean square fluctuation, radius of gyration, and solvent-accessible surface area but the highest number of intra-protein hydrogen bonds among the three structures. Overall, the protein appears to be stable, maintains a compact protein core in nature when not bound to any ligand or any other proteins, and does not change its overall size drastically when in an aqueous solvent. Moreover, Alphafold 2-derived structure also indicated that shorter and stronger hydrogen bonds take place in the protein core or the β-sheets. On the contrary, weak and long hydrogen bonds take place on the loops and α-helices. The protein structure contains 310 helices which are usually dynamic, and intrinsically less stable in nature compared to the α-helices. Loops may shield the hydrophobic core while the loops may interact with ligands and other proteins or potentially regulate the interaction between the core and other molecules. These conclusions set the stage for future wet-lab studies to understand molecular interactions and the structure-function relationship of collagen α1 (XI) during fibrillogenesis.

Acknowledgment

Authors thank colleagues Lisa Warner and Jonathon Reeck for their valuable discussion. Authors acknowledge support from the Institutional Development Awards (IDeA) from the National Institute of General Medical Sciences (#P20GM103408 and P20GM109095) and the BSU - Biomolecular Research Center, RRID:SCR_019174, Lori and Duane Stueckle, and the Idaho State Board of Education.

Contributor Information

Abu Sayeed Chowdhury, Biomolecular Sciences Graduate Program, Center of Biomedical Research Excellence, Boise State University, Boise, Idaho USA.

Julia Thom Oxford, Department of Biological Sciences, Center of Biomedical Research Excellence, Boise State University, Boise, Idaho, USA.

References

  • [1].Rich A and Crick FHC, “The molecular structure of collagen,” Journal of Molecular Biology, vol. 3, no. 5, 1961. [DOI] [PubMed] [Google Scholar]
  • [2].Fraser RDB, MacRae TP, and Suzuki E, “Chain conformation in the collagen molecule,” Journal of Molecular Biology, vol. 129, no. 3, pp. 463–481, 1979. [DOI] [PubMed] [Google Scholar]
  • [3].Morris NP and Bächinger HP, “Type XI collagen is a heterotrimer with the composition (1 alpha, 2 alpha, 3 alpha) retaining non-triple-helical domains.,” Journal of Biological Chemistry, vol. 262, no. 23, pp. 11345–11350, 1987. [PubMed] [Google Scholar]
  • [4].Keene DR, Oxford JT, and Morris NP, “Ultrastructural localization of collagen types II, IX, and XI in the growth plate of human rib and fetal bovine epiphyseal cartilage: Type XI collagen is restricted to thin fibrils.,” Journal of Histochemistry & Cytochemistry, vol. 43, no. 10, pp. 967–979, 1995. [DOI] [PubMed] [Google Scholar]
  • [5].Li Y, Lacerda DA, Warman ML, Beier DR, Yoshioka H, Ninomiya Y, Oxford JT, Morris NP, Andrikopoulos K, Ramirez F, Wardell BB, Lifferth GD, Teuscher C, Woodward SR, Taylor BA, Seegmiller RE, and Olsen BR, “A fibrillar collagen gene, COL11A1, is essential for skeletal morphogenesis,” Cell, vol. 80, no. 3, pp. 423–430, 1995. [DOI] [PubMed] [Google Scholar]
  • [6].Gregory KE, Oxford JT, Chen Y, Gambee JE, Gygi SP, Aebersold R, Neame PJ, Mechling DE, Bächinger HP, and Morris NP, “Structural organization of distinct domains within the non-collagenous N-terminal region of collagen type XI,” Journal of Biological Chemistry, vol. 275, no. 15, pp. 11498–11506, 2000. [DOI] [PubMed] [Google Scholar]
  • [7].Bork P, Downing AK, Kieffer B, and Campbell ID, “Structure and distribution of modules in extracellular proteins,” Quarterly Reviews of Biophysics, vol. 29, no. 2, pp. 119–167, 1996. [DOI] [PubMed] [Google Scholar]
  • [8].MAYNE R and BREWTON R, “New members of the Collagen superfamily,” Current Opinion in Cell Biology, vol. 5, no. 5, pp. 883–890, 1993. [DOI] [PubMed] [Google Scholar]
  • [9].Bork P, “The modular architecture of vertebrate collagens,” FEBS Letters, vol. 307, no. 1, pp. 49–54, 1992. [DOI] [PubMed] [Google Scholar]
  • [10].Leppaönen V-M, Tossavainen H, Permi P, Lehtioö L, Roönnholm G, Goldman A, Kilpelaiönen I, and Pihlajamaa T, “Crystal structure of the N-terminal NC4 domain of collagen IX, a zinc binding member of the laminin-neurexin-sex hormone binding globulin (LNS) domain family,” Journal of Biological Chemistry, vol. 282, no. 32, pp. 23219–23230, 2007. [DOI] [PubMed] [Google Scholar]
  • [11].Fallahi A, Kroll B, Warner LR, Oxford RJ, Irwin KM, Mercer LM, Shadle SE, and Oxford JT, “Structural model of the amino propeptide of collagen XI α1 chain with similarity to the LNS domains,” Protein Science, vol. 14, no. 6, pp. 1526–1537, 2009. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [12].Blaschke UK, Eikenberry EF, Hulmes DJS, Galla H-J, and Bruckner P, “Collagen XI nucleates self-assembly and limits lateral growth of cartilage fibrils,” Journal of Biological Chemistry, vol. 275, no. 14, pp. 10370–10378, 2000. [DOI] [PubMed] [Google Scholar]
  • [13].Medeck RJ, Sosa S, Morris N, Oxford JT. BMP-1-mediated proteolytic processing of alternatively spliced isoforms of collagen type XI. Biochem J. 2003;376(Pt 2):361–368. doi: 10.1042/BJ20030894. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [14].Tan K, Duquette M, huan Liu J, Zhang R, Joachimiak A, huai Wang J, and Lawler J, “The structures of the thrombospondin-1 N-terminal domain and its complex with a synthetic pentameric heparin,” Structure, vol. 14, no. 1, pp. 33–42, 2006. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [15].Pihlajamaa T, Lankinen H, Ylöstalo J, Valmu L, Jäälinoja J, Zaucke F, Spitznagel L, Gösling S, Puustinen A, Mörgelin M, Peränen J, Maurer P, Ala-Kokko L, and Kilpelaïnen I, “Characterization of recombinant amino-terminal NC4 domain of human collagen IX,” Journal of Biological Chemistry, vol. 279, no. 23, pp. 24265–24273, 2004. [DOI] [PubMed] [Google Scholar]
  • [16].Webb B and Sali A, “Comparative protein structure modeling using modeller,” Current Protocols in Protein Science, vol. 86, no. 1, 2016. [DOI] [PubMed] [Google Scholar]
  • [17].McDougal OM, Warner LR, Mallory C, Oxford JT, “PREDICTED STRUCTURE AND BINDING MOTIFS OF COLLAGEN α1(XI),” GSTF Int J Bioinforma Biotechnol, vol 1(1), pp. 43–48, 2011. [PMC free article] [PubMed] [Google Scholar]
  • [18].Tisi D, “Structure of the C-terminal laminin G-like domain pair of the laminin alpha 2 chain harbouring binding sites for alpha -dystroglycan and heparin,” The EMBO Journal, vol. 19, no. 7, pp. 1432–1440, 2000. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [19].Jumper J, Evans R, Pritzel A et al. , “Highly accurate protein structure prediction with AlphaFold. Nature,” vol 596, pp. 583–589, 2021. 10.1038/s41586-021-03819-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [20].Berendsen HJC, van der Spoel D, and van Drunen R, “Gromacs: A message-passing parallel molecular dynamics implementation,” Computer Physics Communications, vol. 91, no. 1–3, pp. 43–56, 1995. [Google Scholar]
  • [21].Pettersen EF, Goddard TD, Huang CC, Meng EC, Couch GS, Croll TI, Morris JH, and Ferrin TE, ‘UCSF chimerax : Structure visualization for researchers, educators, and developers,” Protein Science, vol. 30, no. 1, pp. 70–82, 2020. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [22].Bystroff C and Shao Y, “Fully automated ab initio protein structure prediction using I-sites, HMMSTR and Rosetta,” Bioinformatics, vol. 18, no. Suppl 1, 2002. [DOI] [PubMed] [Google Scholar]
  • [23].Ramachandran S, Kota P, Ding F, and Dokholyan NV, “Automated minimization of steric clashes in protein structures,” Proteins: Structure, Function, and Bioinformatics, vol. 79, no. 1, pp. 261–270, 2010. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [24].Martí-Renom MA, Stuart AC, Fiser A, Sánchez R, Melo F, and Šali A, “Comparative protein structure modeling of genes and genomes,” Annual Review of Biophysics and Biomolecular Structure, vol. 29, no. 1, pp. 291–325, 2000. [DOI] [PubMed] [Google Scholar]
  • [25].Jorgensen WL, Maxwell DS, and Tirado-Rives J, “Development and testing of the OPLS all-atom force field on conformational energetics and properties of organic liquids,” Journal of the American Chemical Society, vol. 118, no. 45, pp. 11225–11236, 1996. [Google Scholar]
  • [26].Tan K, Duquette M, J.-huan Liu, Zhang R, Joachimiak A, J.-huai Wang, and Lawler J, ‘The structures of the thrombospondin-1 N-terminal domain and its complex with a synthetic pentameric heparin,” Structure, vol. 14, no. 1, pp. 33–42, 2006. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [27].Grishkovskaya I, Avvakumov GV, Hammond GL, and Muller YA, “Resolution of a disordered region at the entrance of the human sex hormone-binding globulin steroid-binding site,” Journal of Molecular Biology, vol. 318, no. 3, pp. 621–626, 2002. [DOI] [PubMed] [Google Scholar]
  • [28].Stevens AO and He Y, “Benchmarking the accuracy of alphafold 2 in loop structure prediction,” Biomolecules, vol. 12, no. 7, p. 985, 2022. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [29].Vieira-Pires RS and Morais-Cabral JH, “310 helices in channels and other membrane proteins,” Journal of General Physiology, vol. 136, no. 6, pp. 585–592, 2010. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [30].Jiang Z, You L, Dou W, Sun T, and Xu P, “Effects of an electric field on the conformational transition of the protein: A molecular dynamics simulation study,” Polymers, vol. 11, no. 2, p. 282, 2019. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [31].Panja AS, Maiti S, and Bandyopadhyay B, “Protein stability governed by its structural plasticity is inferred by physicochemical factors and Salt Bridges,” Scientific Reports, vol. 10, no. 1, 2020. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [32].Lins L, Thomas A, and Brasseur R, “Analysis of accessible surface of residues in proteins,” Protein Science, vol. 12, no. 7, pp. 1406–1417, 2003. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [33].Y.-wei Dong, M.-ling Liao, liangMeng X, and Somero GN, “Structural flexibility and protein adaptation to temperature: Molecular dynamics analysis of Malate dehydrogenases of marine molluscs,” Proceedings of the National Academy of Sciences, vol. 115, no. 6, pp. 1274–1279, 2018. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [34].Papaleo E, Saladino G, Lambrughi M, Lindorff-Larsen K, Gervasio FL, and Nussinov R, ‘The role of protein loops and linkers in conformational dynamics and allostery,” Chemical Reviews, vol. 116, no. 11, pp. 6391–6423,2016. [DOI] [PubMed] [Google Scholar]
  • [35].Wu S-J and Dean DH, “Functional significance of loops in the receptor binding domain ofBacillus thuringiensisCryIIIA δ-endotoxin,” Journal of Molecular Biology, vol. 255, no. 4, pp. 628–640, 1996. [DOI] [PubMed] [Google Scholar]
  • [36].Shi L and Javitch JA, “The second extracellular loop of the dopamine D 2 receptor lines the binding-site Crevice,” Proceedings of the National Academy of Sciences, vol. 101, no. 2, pp. 440–445, 2004. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [37].Jones S and Thornton JM, “Prediction of protein-protein interaction sites using patch analysis 1 1edited by Von Heijne G,” Journal of Molecular Biology, vol. 272, no. 1, pp. 133–143, 1997. [DOI] [PubMed] [Google Scholar]
  • [38].Herschlag D and Pinney ΜM, ‘Hydrogen bonds: Simple after all?,” Biochemistry, vol. 57, no. 24, pp. 3338–3352, 2018. [DOI] [PubMed] [Google Scholar]
  • [39].Gasteiger E, “Expasy: The Proteomics Server for in-depth protein knowledge and analysis,” Nucleic Acids Research, vol. 31, no. 13, pp. 3784–3788, 2003. [DOI] [PMC free article] [PubMed] [Google Scholar]

RESOURCES