LOCPRED

LOCPRED predicts local protein 3D structure from amino acid sequences using a Protein Blocks structural alphabet derived from PDB statistics and a Bayesian mapping of sequence windows to Protein Blocks.

Key Features:

• Protein Block Alphabet: Utilizes a set of 16 Protein Blocks derived from statistical analyses of Protein Data Bank (PDB) structures.
• Dihedral Angle Characterization: Represents each Protein Block by the backbone dihedral angles (φ, ψ) of five consecutive amino acid residues.
• Sequence Window Amino Acid Distributions: Leverages observed amino acid distributions within sequence windows that encompass Protein Blocks.
• Bayesian Prediction Methodology: Applies a Bayesian approach to assign Protein Blocks and predict local 3D structure from sequence information.

Scientific Applications:

• Structural Biology: Infers local 3D structural motifs to aid interpretation of protein structures and models.
• Computational Biochemistry: Provides local structural annotations that can inform computational analyses of protein behavior.
• Molecular Dynamics: Supplies local structural prototypes useful for initializing or interpreting molecular dynamics simulations.
• Protein Folding Analysis: Assists in elucidating local folding mechanisms by mapping sequence to local structural states.
• Protein Function Inference: Supports functional hypotheses by relating local structure to potential functional sites.
• Protein Design: Enables design efforts by predicting local structural propensities from sequence variants.
• Drug Discovery: Contributes local structural information that can inform target characterization and ligand interaction hypotheses.
• Evolutionary Studies: Facilitates comparative analyses by mapping sequence variation to conserved or variable local structural elements.

Methodology:

Derives 16 Protein Blocks from statistical analysis of PDB structures; characterizes each block by φ, ψ angles of five-residue fragments; measures amino acid distributions within sequence windows around blocks; and uses a Bayesian algorithm to predict Protein Block assignments from sequence.