Missense3D-DB

Missense3D-DB provides precomputed structural interpretations of approximately 3,960,015 human protein-coding missense variants using three-dimensional structural information from experimental coordinates and in-house predicted models to assess variant impacts on protein structure.

Key Features:

• Atom-Based Mapping: Applies a large-scale atom-based mapping approach to analyze missense variants mapped onto 18,874 experimental structures and 84,818 predicted models.
• Variant Sources: Integrates variant data from GnomAD, ClinVar, and UniProt for structural analysis of 3,960,015 missense variants.
• Annotation Integration: Incorporates population frequency data and functional predictions from SIFT and PolyPhen alongside structural annotations.
• Structural Impact Estimates: Predicts how amino acid substitutions affect protein structure, reporting that approximately 14% of structurally analyzable variants from GnomAD may influence protein structure.
• Missense3D-PPI Extension: Targets protein-protein interaction (PPI) sites to detect damaging surface variants that disrupt PPI interfaces.
• Performance Comparison: Demonstrates higher sensitivity and accuracy than its predecessor and than BeAtMuSiC, mCSM-PPI2, and MutaBind2 for predicting PPI impacts, with lower specificity than Missense3D.

Scientific Applications:

• Variant Prioritization: Aids prioritization of variants of unknown clinical significance and variants with conflicting interpretations by providing structural impact evidence.
• Molecular Mechanism Research: Supports investigation of molecular mechanisms underlying genetic disorders through residue-level structural analyses.
• Therapeutic Development: Informs development of targeted therapeutic strategies by identifying structural perturbations caused by missense variants.

Methodology:

Precomputed atom-based mapping of missense variants onto three-dimensional structures derived from experimental coordinates and in-house predicted models, integration of annotations from GnomAD, ClinVar, UniProt and predictions from SIFT and PolyPhen, and a PPI-targeting Missense3D-PPI extension evaluated against BeAtMuSiC, mCSM-PPI2, and MutaBind2.