"Adapted Linear Interaction Energy": A Structure-Based LIE Parametrization for Fast Prediction of Protein-Ligand...

Area of Science:

• Computational chemistry
• Molecular modeling
• Drug discovery

Background:

• Predicting protein-ligand binding affinity is crucial for drug discovery.
• Existing Linear Interaction Energy (LIE) methods require empirical parameters.
• A need exists for more accurate and generalizable binding energy prediction models.

Purpose of the Study:

• To develop a structure-based parametrization of the LIE method.
• To introduce the Adapted LIE (ALIE) model for predicting absolute protein-ligand binding energies.
• To evaluate ALIE's performance against standard LIE and LIE + γSASA models.

Main Methods:

• Developed a structure-based parametrization for the LIE method.
• Defined system-dependent descriptors for α and β coefficients in ALIE.
• Tested ALIE on diverse in-house and external datasets.
• Compared ALIE with standard LIE and LIE + γSASA models.

Main Results:

• The best ALIE formulation achieved a mean average deviation of 1.8 kcal/mol.
• ALIE requires only one fitted parameter and no empirical γ term.
• The model demonstrated robustness against additional fitting and cross-validation.
• ALIE showed improved performance compared to standard LIE and LIE + γSASA.

Conclusions:

• ALIE provides an accurate and robust method for predicting absolute protein-ligand binding energies.
• The structure-based parametrization eliminates the need for empirical terms, enhancing generalizability.
• ALIE represents a significant advancement in computational drug design and molecular modeling.