Chimera visualisation plugin

Chimera visualisation plugin detects outlier residues in cryo-electron microscopy (cryo-EM) derived protein models using an unsupervised probabilistic anomaly detection approach coupled to a histogram-based outlier score (HBOS) derived from high-resolution X-ray reference data.

Key Features:

• Unsupervised outlier detection: Uses an unsupervised machine learning probabilistic anomaly detection model that does not require labeled data to identify anomalous residues.
• Histogram-Based Outlier Score (HBOS): Employs an HBOS computed from a high-resolution X-ray dataset (resolution <1.5 Å) as the reference standard for scoring residues.
• Structural parameters analyzed: Stores and analyzes distal block distance, side-chain length, phi and psi backbone angles, and the first chi side-chain angle for each residue.
• HBOS database: Maintains a database of structural parameter histograms derived from the high-resolution X-ray reference dataset for comparison against cryo-EM models.
• Outlier thresholding: Flags residues with HBOS values exceeding 10 as outliers.

Scientific Applications:

• Cryo-EM model validation: Identifies residue-level outliers in cryo-EM derived protein models to support validation efforts.
• Structural model refinement: Highlights specific residues for targeted refinement or re-modeling during structure determination workflows.
• Biological interpretation support: Improves reliability of structural interpretations of biological function and molecular interactions by identifying potential modeling errors.

Methodology:

Structural parameters are collected from a high-resolution X-ray reference dataset (resolution <1.5 Å) to build an HBOS database, and an unsupervised probabilistic anomaly detection model evaluates cryo-EM derived protein models by computing HBOS per residue and flagging values >10.