Computational Design of PDZ-Peptide Binding.

Area of Science:

• Biochemistry
• Computational Biology
• Structural Biology

Background:

• PDZ domains are crucial protein interaction modules involved in various cellular processes.
• Understanding PDZ-peptide interactions is vital for drug discovery and understanding cellular signaling.

Purpose of the Study:

• To present and validate two computational methods for predicting PDZ-peptide binding.
• To introduce a new high-throughput computational protein design (CPD) method for PDZ-peptide interactions.
• To assess the accuracy of a medium-throughput approach combining molecular dynamics (MD) and Poisson-Boltzmann (PB) methods.

Main Methods:

• A novel CPD method utilizing adaptive Monte Carlo simulations for efficient sampling of peptide variants and precise binding free energy estimation.
• A medium-throughput approach combining MD for conformational sampling and PB Linear Interaction Energy for scoring.
• Detailed protocol provided for CPD using Proteus software; MD/PB methods applicable with NAMD and Charmm.

Main Results:

• The new CPD method efficiently samples peptide variants and estimates relative binding free energies.
• The medium-throughput MD/PB approach achieved high accuracy for 40 Tiam1-peptide complexes, with mean errors of ~0.5 kcal/mol.
• No significant errors were observed in the relative binding free energy predictions using the medium-throughput method.

Conclusions:

• The presented computational methods provide accurate predictions for PDZ-peptide binding free energies.
• The novel CPD approach offers an efficient way to design peptides with high affinity for PDZ domains.
• The medium-throughput MD/PB approach is accurate but requires parameter fitting and its transferability to other protein families is yet to be determined.