Dictionary of protein secondary structure: Pattern recognition of hydrogen‐bonded and geometrical features

By W. Kabsch, C. Sander. 1983

For a successful analysis of the relation between amino acid sequence and protein structure, an unambiguous and physically meaningful definition of secondary structure is essential. We have developed a set of simple and physically motivated criteria for secondary structure, programmed as a pattern‐recognition process of hydrogen‐bonded and geometrical features extracted from x‐ray coordinates. Cooperative secondary structure is recognized as repeats of the elementary hydrogen‐bonding patterns “turn” and “bridge.” Repeating turns are “helices,” repeating bridges are “ladders,” connected ladders are “sheets.” Geometric structure is defined in terms of the concepts torsion and curvature of differential geometry. Local chain “chirality” is the torsional handedness of four consecutive Cα positions and is positive for right‐handed helices and negative for ideal twisted β‐sheets. Curved pieces are defined as “bends.” Solvent “exposure” is given as the number of water molecules in possible contact with a residue. The end result is a compilation of the primary structure, including SS bonds, secondary structure, and solvent exposure of 62 different globular proteins. The presentation is in linear form: strip graphs for an overall view and strip tables for the details of each of 10.925 residues. The dictionary is also available in computer‐readable form for protein structure prediction work.

Subfields: Computational Biology & Bioinformatics

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computational Biology & Bioinformatics

1. Jinbo Xu
2. Yang Zhang
3. Sheng Wang
4. David T. Jones
5. Björn Wallner
6. David Baker
7. Burkhard Rost
8. Arne Elofsson

Citations in the corpus

• Critical assessment of protein intrinsic disorder prediction (CAID) ‐ Results of round 2 (2023, background): “Nota-bly, AlphaFold-binding, which prioritizes regions with high structural propensity (pLDDT) and high relative solvent accessibility (RSA), has the best F max and a good APS but falls in the 13th position when considering AUC.”
• Protein Sequence Analysis Using the MPI Bioinformatics Toolkit (2020, methodology): “Like query HMMs, database HMMs include secondary structure information, either predicted by PSIPRED or assigned based on 3D structure by DSSP (Joosten et al., 2011; Kabsch & Sander, 1983).”
• Improved protein structure refinement guided by deep learning based accuracy estimation (2020, positive): “Refinement improved models across various secondary structures to similar extents, and corrected secondary structures originally modeled incorrectly, increasing model secondary structure accuracy by almost 10% based on an 8-state definition ​(25)​ (Figure 4E).”
• HH-suite3 for fast remote homology detection and deep protein annotation (2019, methodology): “pl, and we added structurally defined secondary structure states from DSSP [36] using addss.”
• The SWISS-MODEL Repository—new features and functionality (2016, methodology): “Additionally, each entry is annotated with a sequence profile (HMM (16)), predicted secondary structure via SSpro (17) and PSIPRED (18), secondary structure via DSSP (19), predicted solvent accessibility via ACCpro (17), and per residue solvent accessibility via NACCESS (S. Hubbard and J.M. Thornton”
• Toward the estimation of the absolute quality of individual protein structure models (2010, methodology): “The secondary structure and solvent accessibility of the model are calculated by DSSP (Kabsch and Sander, 1983).”
• Modeling aspects of the language of life through transfer-learning protein sequences (2019, background): “DSSP [68] assigned secondary structure and intrinsically disordered residues are flagged (residues without atomic coordinates, i.e. REMARK-465 in the PDB file).”
• NetSurfP-2.0: improved prediction of protein structural features by integrated deep learning (2018, methodology): “For all residues in each chain in the training dataset, we calculated its absolute and relative solvent accessibility (ASA and RSA, respectively), 3- and 8-class secondary structure classification (SS3 and SS8, respectively), and the backbone dihedral angles φ and ψ using the DSSP software.(17) Fina”
• Quantitative missense variant effect prediction using large-scale mutagenesis data (2017, positive): “To obtain structural information, we use DSSP (http://www.cmbi.ru.nl/dssp.html) (Kabsch and Sander, 1983) and PDB files from the Protein Data Bank (http://www.rcsb.org/pdb/home/home.do) (Rose et al., 2011).”
• Deep Supervised and Convolutional Generative Stochastic Network for Protein Secondary Structure Prediction (2014, methodology): “8-states secondary structure labels and solvent accessibility score were inferred from the 3D PDB structure by the DSSP program (Kabsch & Sander, 1983).”
• Protein sequence design with a learned potential (2022, methodology): “Additionally, we restrict the model from designing glycines at non-loop positions, based on DSSP assignment 53,54 .”
• TM-align: a protein structure alignment algorithm based on the TM-score (2005, neutral): “We note that the set of eight parameters are optimized based on 100 non-homologous training proteins by maximizing the SS assignment similarity to the DSSP definition (20), which defines protein SS elements on the basis of hydrogen bond patterns and requires the full set of backbone atomic coordinat”