IDPConformerGenerator

IDPConformerGenerator generates large, diverse ensembles of intrinsically disordered protein (IDP) conformations by sampling backbone torsion angles and constructing side chains to represent IDP structural ensembles for validation against experimental data.

Key Features:

• Modular Design: Modular architecture exposes configurable components for tailoring conformer sampling and modeling.
• Conformer Generation: Constructs conformers by sampling backbone torsion angles—phi (φ), psi (ψ), and omega (ω)—from sequence fragments derived from loops and secondary structure elements found in folded protein structures within the RCSB Protein Data Bank.
• Side Chain Building: Builds side chains using robust Monte Carlo algorithms and expanded rotamer libraries to enhance accuracy and diversity.
• Physical Restraints: Ensembles adhere to geometric, steric, and other physical restraints based on input sequences.
• User-Defined Options: Allows specification of variables such as fractional sampling of secondary structures.
• Bayesian Models: Implements Bayesian models to evaluate the consistency of IDP ensembles with experimental data.
• Machine Learning Integration: Incorporates a machine learning approach to transform between internal and Cartesian coordinates to reduce spatial representation errors.

Scientific Applications:

• Ensemble Generation: Generate comprehensive conformational ensembles representing IDP heterogeneity for structural characterization.
• Experimental Validation: Assess and validate IDP ensembles against experimental measurements using Bayesian evaluation.
• Functional and Mechanistic Insight: Provide structural insights into IDP mechanisms, functions, and roles in biological processes and disease.

Methodology:

Extracts sequence fragments from known protein structures and samples backbone torsion angles (φ, ψ, ω) to build conformers; builds side chains with Monte Carlo algorithms and expanded rotamer libraries; applies geometric, steric, and other physical restraints; evaluates ensembles with Bayesian models; and employs machine learning to convert between internal and Cartesian coordinates.