RPBS

RPBS predicts local protein 3D structure from amino acid sequence using a Protein Blocks structural alphabet to enable analysis and modeling of local backbone conformations.

Key Features:

• Protein Blocks (PBs): Uses a structural alphabet of 16 Protein Blocks defined by phi (φ) and psi (ψ) dihedral angles over five consecutive residues, each representing a small 3D prototype.
• Bayesian local structure prediction: Employs a Bayesian approach that predicts PBs from sequence by analyzing amino acid distributions within sequence windows.
• Integration with PDB statistics: Incorporates statistical analyses of Protein Data Bank (PDB) structures to refine amino acid distribution models for PBs.
• Loop clustering: Clusters loops into statistically meaningful families based on backbone structures to support loop classification and modeling.
• Automatic loop modeling: Assigns query loops to families based on sequence and flank backbone structures, explores conformational space for target loop backbones, and performs energy minimization for loop closure.
• Reliability scoring: Provides statistical reliability scores for predictions that correlate with model quality measured by root-mean-square deviation (rmsd).
• Sequence similarity searches with deterministic automatons: Supports detection of shared oligopeptide segments across large protein databases using deterministic finite state automatons analogous to BLAST and Automat.
• Graphical and textual outputs: Produces predictions in both textual and graphical formats for structural interpretation.

Scientific Applications:

• Protein modeling: Predicts loop regions between secondary structures and facilitates modeling by assigning loops to backbone-based families.
• Sequence similarity and comparative analysis: Identifies shared oligopeptide segments across databases to support comparative and evolutionary studies.
• Viral and autoimmune research: Supports analysis of viral proteins such as HIV-1 and autoimmune disease–related sequences by identifying sequence similarities and structural motifs.
• Modeling of ambiguous loops: Enables modeling of loops that are not precisely delimited in sequence data.

Methodology:

Defines 16 Protein Blocks from phi/psi dihedral angles over five-residue fragments; applies a Bayesian prediction framework using amino acid distributions within sequence windows; integrates statistical analyses of PDB structures; uses deterministic finite state automatons for sequence similarity searches; assigns loops to families and models loops by exploring conformational space with energy minimization and assigns reliability scores correlated with rmsd.