CPPCGM: A Highly Efficient Sequence-Based Tool for Simultaneously Identifying and Generating Cell-Penetrating...

Area of Science:

• Biochemistry
• Computational Biology
• Drug Delivery Systems

Background:

• Cell-penetrating peptides (CPPs) are crucial for intracellular drug delivery but experimental identification is costly and time-consuming.
• Existing computational methods for CPP screening have limitations in feature representation, hindering performance.
• Protein language models (PLMs) offer advanced capabilities for peptide analysis and design.

Purpose of the Study:

• To develop a novel deep learning framework, CPPCGM, for identifying and generating cell-penetrating peptides (CPPs).
• To overcome the limitations of current computational methods in feature representation for CPP discovery.
• To leverage PLMs for efficient and accurate CPP candidate screening and novel peptide generation.

Main Methods:

• Developed CPPCGM, a deep learning framework comprising a CPPClassifier and a CPPGenerator.
• The CPPClassifier utilizes three pre-trained protein language models for robust CPP/non-CPP classification.
• The CPPGenerator, inspired by generative adversarial networks, creates novel peptide sequences not present in the training data.

Main Results:

• CPPCGM achieved high classification performance with Matthews correlation coefficient scores of 0.876, 0.923, and 0.664 on three independent datasets.
• The framework significantly outperformed existing state-of-the-art methods in CPP identification.
• Qualitative and quantitative evaluations confirmed the successful generation of novel, potentially functional CPPs by the CPPGenerator.

Conclusions:

• The CPPCGM framework demonstrates superior performance in both classifying and generating cell-penetrating peptides.
• Utilizing protein language models within CPPCGM significantly enhances the efficiency and accuracy of CPP discovery.
• This approach holds promise for optimizing peptides for biochemical functions, advancing drug delivery and biomedical applications.