MODELLER

MODELLER constructs three-dimensional protein models from sequence using comparative (homology) modeling to predict atomic-resolution protein structures when experimental structures are unavailable.

Key Features:

• Comparative modeling: Predicts target 3D structures by aligning a target sequence to one or more template structures and performing fold assignment, target-template alignment, model building, and model evaluation.
• Spatial restraints optimization: Uses probability density functions (pdfs) for structural features—Cα-Cα distances, main-chain N-O distances, and main- and side-chain dihedral angles—and optimizes positions of all non-hydrogen atoms via a variable target function and the conjugate gradients algorithm.
• Loop modeling: Optimizes loop regions by adjusting all non-hydrogen atoms within a fixed environment using a pseudo energy function that combines CHARMM-22 spatial restraints and statistical dihedral and atomic contact preferences.
• Automation and scalability: Automates the comparative modeling workflow to enable large-scale and genome-wide structure prediction applications.
• Integration with genomic data: Integrates comparative modeling with genome sequencing and functional genomics data for automated, high-throughput structural annotation.

Scientific Applications:

• Structure prediction: Predicts protein structures in cases where experimental atomic-resolution data are lacking.
• Genomic-scale modeling: Models proteins from organisms with uncharacterized genomes and supports genome-wide structural annotation.
• Functional genomics: Provides structural models to support interpretation of gene products in functional genomics projects.
• Protein function and interaction analysis: Supplies 3D models to aid understanding of protein function and interaction networks.

Methodology:

Fold assignment; target-template alignment; model building by satisfying spatial restraints expressed as probability density functions (pdfs) for Cα-Cα distances, main-chain N-O distances and main- and side-chain dihedral angles; optimization of all non-hydrogen atom positions using a variable target function and the conjugate gradients algorithm; loop modeling using a pseudo energy function combining CHARMM-22 spatial restraints and statistical dihedral/contact preferences; model evaluation.