clmDCA

clmDCA predicts inter-residue contacts in proteins to inform protein tertiary structure determination.

Key Features:

• Composite Likelihood Maximization: Uses composite likelihood (the product of conditional probabilities of all residue pairs) instead of pseudo-likelihood for parameter estimation in Markov Random Field models.
• Efficiency and Accuracy: Balances computational efficiency with improved prediction accuracy and demonstrates superior performance relative to existing MRF-based approaches on PSICOV and CASP-11 benchmarks.
• Integration with Deep Learning: Supports refinement of predicted contacts through integration with deep learning techniques.
• Successful Applications: Predicted contacts have been applied to construct tertiary structures of proteins in the PSICOV dataset.

Scientific Applications:

• Protein Structure Prediction: Provides inter-residue contact predictions used to assist determination of protein tertiary structures.
• Structural Biology Research: Supplies contact information to study protein folding, stability, and interaction networks.

Methodology:

Applies composite likelihood maximization for parameter estimation in Markov Random Fields, where composite likelihood is defined as the product of conditional probabilities of residue pairs; integrates deep learning for prediction refinement and has been evaluated on PSICOV and CASP-11.