Development and Experimental Validation of a Machine-Learning- and Physics- Based Exhaustive Hexapeptide Screening...

Area of Science:

• Biopharmaceutical research
• Computational drug discovery
• Artificial intelligence in medicine

Background:

• Peptide therapeutics are a growing area in biopharmaceuticals.
• Current experimental screening methods are costly and slow.
• Efficient pipelines are needed for peptide drug discovery.

Purpose of the Study:

• To develop an AI-driven framework for efficient hexapeptide screening.
• To integrate physical interaction data with sequence features for improved accuracy.
• To accelerate the discovery of novel peptide therapeutics.

Main Methods:

• A two-stage AI pipeline was designed for exhaustive hexapeptide screening.
• Peptide-protein complexes were encoded using a fragmentation-based representation.
• Docking-derived binding energies were combined with a transformer model to learn sequence-energy relationships.

Main Results:

• The AI framework achieved high hit rates in screening.
• Novel hexapeptides with micromolar binding affinities were identified for NRP-1, STING, and cGAS.
• The framework demonstrated robustness and generalizability across different targets.

Conclusions:

• The AI framework effectively bridges physical modeling and feature representation for peptide discovery.
• This approach significantly accelerates the identification of potential peptide drug candidates.
• The developed method offers a powerful tool for the biopharmaceutical industry.