Dipcheck

Dipcheck evaluates protein backbone geometry using a three-dimensional distance-geometry metric (DipSpace) to validate and characterize five-atom dipeptide conformations for structural-model validation.

Key Features:

• DipSpace Framework: A Euclidean 3D space constructed from orthogonal descriptors that characterize the geometry of the five-atom dipeptide unit (CAi-1-Oi-1-CAi-Oi-CAi+1), providing a probabilistic description of backbone geometry.
• Comprehensive Database: A reference database comprising approximately 1,024,000 data points derived from well-refined Protein Data Bank (PDB) structures.
• Metric Calculation: Metrics computed from eigenvalues of Euclidean distance matrices for dipeptide units using ~1.3 million nonredundant, high-quality dipeptide units and principal component analysis to define DipSpace.
• Geometric Descriptors: Three DipSpace axes that describe local extension, twist, and bend to identify Cα atoms in unusual or unlikely geometrical environments.
• Validation Capabilities: Detection of local and overall backbone deviations and conserved unusual conformations within protein families (e.g., trypsin proteases), indicating geometrically strained residues with potential functional significance.

Scientific Applications:

• Structural-model validation: Identification of backbone geometrical deviations for validating protein models derived from crystallography or other structural methods.
• Functional inference: Pinpointing geometrically strained residues and conserved unusual conformations that may have functional significance, as observed in trypsin proteases.
• Conformational analysis: Quantitative characterization of dipeptide conformational space for studies of protein dynamics and interactions.

Methodology:

Compute Euclidean distance matrices for five-atom dipeptide units, calculate eigenvalues of these matrices, assemble PDB-derived dipeptide datasets (~1.3 million nonredundant units; ~1,024,000 reference points), and apply principal component analysis to define and populate DipSpace.