A Transferable Coarse Grain Non-bonded Interaction Model For Amino Acids.

Area of Science:

• Computational Biology
• Biophysics
• Structural Bioinformatics

Background:

• Genomic data generation outpaces experimental structure determination, necessitating accurate protein structure prediction tools.
• Coarse-grained (CG) models are vital for computational protein structure prediction, offering high performance.
• Developing generalized, high-quality CG models remains a significant challenge.

Purpose of the Study:

• To present a novel coarse-grained (CG) based interaction potential for naturally occurring amino acids.
• To parameterize a CG model using experimental thermodynamic data, distinct from all-atom (AA) simulations or experimental structures.
• To develop a dataset for modeling peptides and proteins using this novel potential.

Main Methods:

• Condensing three to four heavy atoms and hydrogens into single CG sites.
• Parameterizing site-site interaction potentials using experimental thermodynamic data (surface tension, density).
• Utilizing Lennard-Jones (LJ) functional forms for intermolecular potentials.

Main Results:

• Developed an amino acid potential dataset for peptide and protein modeling.
• Evaluated the model by comparing solvent accessible surface area (SASA) to AA representations.
• Demonstrated strong performance in ranking protein decoy datasets compared to existing models.

Conclusions:

• The novel CG potential, parameterized by experimental thermodynamic data, shows promise for protein structure prediction.
• This approach offers a new method for developing CG models without relying on AA simulations or experimental structural data.
• The developed potential performs competitively with existing prediction models for key properties like SASA and decoy ranking.