Reformulation of the protein databank for real-time search of geometrical attributes of protein structures

TL;DR

This paper introduces PDBMine, a platform for analyzing protein structures by examining dihedral angles and sequence patterns in the Protein Data Bank.

Contribution

The novel contribution is PDBMine, a scalable platform for real-time mining of protein structure geometrical attributes and sequence dependencies.

Findings

Longer k-mers show significant deviations from statistical independence, indicating context-dependent amino acid co-occurrence.

Increasing sequence context reduces dihedral angle variability, aligning better with observed backbone geometries.

A clustering method identifies dominant structural motifs from full-sequence dihedral conformations.

Abstract

In this study, we introduce the design and implementation of PDBMine, a large-scale, queryable platform for mining sequence-structure statistics from the Protein Data Bank (PDB). PDBMine enables rapid analysis of local conformational trends across proteins by extracting dihedral angles and sequence patterns at scale. In addition to the design and implementation of PDBMine, we also present results validating its ability to return structurally meaningful information.

We first assess the accuracy of its dihedral angle distributions by comparing them to established Ramachandran space and verifying expected behaviors of residues such as glycine and proline. We then use PDBMine to analyze the statistical properties of amino acid subsequences of length
k
= 1 to 5.

Our findings reveal that longer
k
-mers exhibit significant departures from statistical independence, suggesting context-dependent constraints on amino acid co-occurrence. We also show that increasing local sequence context restricts dihedral angle variability, with longer
k
-mers producing distributions that more closely align with experimentally observed backbone geometries. Finally, we present a high-dimensional clustering method for grouping full-sequence dihedral conformations, enabling identification of dominant local structural motifs.

These results highlight PDBMine’s potential as a versatile tool for structure validation, statistical modeling, and probing the principles that govern sequence-structure compatibility in proteins.

Figures