Multicoil2

Multicoil2 predicts coiled-coil regions in amino acid sequences and classifies their oligomerization state, distinguishing parallel and antiparallel dimeric and trimeric alpha-helical assemblies.

Key Features:

• Pairwise Correlations: Utilizes localized pairwise amino-acid correlations to improve prediction accuracy.
• Hidden Markov Models (HMMs): Employs HMMs to model sequence-position variability of coiled-coil segments.
• Markov Random Fields (MRFs): Integrates pairwise correlations and HMM outputs within an MRF framework to model sequence dependencies.
• Multinomial Logistic Regression: Integrates sequence features within the MRF to derive an optimized scoring function.
• Scoring Function: Produces MRF potentials that capture localized pairwise correlations to distinguish dimeric, trimeric, and non-coiled-coil states.
• Training Database: Trained on a database of 2,105 sequences (124,088 residues) with structural annotations derived from experimental data.

Scientific Applications:

• Coiled-coil detection and oligomerization prediction: Identification of coiled-coil regions and classification into dimeric versus trimeric states in protein sequences.
• Protein structural annotation: Assignment of alpha-helical coiled-coil topology (parallel/antiparallel) to amino-acid sequences.
• Study of protein assembly and function: Analysis of coiled-coil-mediated assembly and interaction motifs to investigate protein structure and function.

Methodology:

Combines pairwise correlations with Hidden Markov Models in a Markov Random Field framework, integrates sequence features via multinomial logistic regression to derive an optimized scoring function producing MRF potentials that capture localized pairwise correlations, is trained on a database of 2,105 sequences (124,088 residues) with structural annotations from experimental data, and is validated by leave-family-out cross-validation.