In Silico Screening and Optimization of Cell-Penetrating Peptides Using Deep Learning Methods.

Area of Science:

• Biochemistry
• Bioinformatics
• Molecular Biology

Background:

• Cell-penetrating peptides (CPPs) are crucial for intracellular delivery of bioactive molecules.
• Developing novel and efficient CPPs remains an ongoing challenge in biomedical research.
• In silico methods for CPP discovery often yield high false-positive rates, necessitating improved prediction strategies.

Purpose of the Study:

• To develop a novel deep-learning-based prediction method, AiCPP, for identifying efficient cell-penetrating peptides.
• To reduce false-positive predictions in silico CPP discovery.
• To discover and validate new CPPs with potential therapeutic applications.

Main Methods:

• Development of a deep-learning model (AiCPP) for peptide sequence analysis.
• Utilizing a large dataset of peptide sequences from human-reference proteins as a negative set to minimize false positives.
• Training the model to recognize short peptide sequence motifs indicative of CPP activity.

Main Results:

• AiCPP successfully identified short peptide sequences derived from amyloid precursor proteins as efficient novel CPPs.
• Experimental validation confirmed the cell-penetrating capabilities of these newly discovered CPP sequences.
• The study demonstrated the potential for further optimization of these identified CPP sequences.

Conclusions:

• The deep-learning approach (AiCPP) offers a more accurate and efficient method for discovering cell-penetrating peptides.
• Novel CPPs derived from amyloid precursor proteins show promise for drug delivery applications.
• AiCPP facilitates the development of next-generation CPPs by improving in silico prediction accuracy.