Computational peptide discovery with a genetic programming approach.

Area of Science:

• Protein engineering
• Computational biology
• Biomarker discovery

Background:

• Developing therapeutic peptides and diagnostic biomarkers is complex and time-consuming.
• Current methods require extensive lab work and struggle with vast search spaces.
• In silico approaches, like evolutionary algorithms, can accelerate peptide discovery and reduce costs.

Purpose of the Study:

• To develop and utilize a novel variant of the POET algorithm, POET$$, for efficient peptide discovery.
• To generate peptides that enhance the sensitivity of chemical exchange saturation transfer (CEST) in magnetic resonance imaging (MRI).
• To explore the potential of combining genetic programming with regular expressions for identifying novel peptide targets.

Main Methods:

• Developed POET$$, a genetic programming algorithm representing individuals as regular expressions.
• Trained the algorithm on a curated dataset.
• Applied the trained model to generate peptides for improving CEST-MRI sensitivity.

Main Results:

• The POET$$ model demonstrated a 20% performance gain over initial POET models.
• A candidate peptide was identified with a 58% performance increase compared to the gold-standard.
• The algorithm successfully identified new peptide targets enhancing CEST detection sensitivity.

Conclusions:

• POET$$ offers an efficient in silico approach for identifying peptides with therapeutic or diagnostic potential.
• The combination of genetic programming and regular expressions accelerates the discovery of sensitive peptides for CEST-MRI.
• This method presents a promising direction for advancing protein engineering and disease diagnostics.