POSEIDON: Peptidic Objects SEquence-based Interaction with cellular DOmaiNs: a new database and predictor.

Area of Science:

• Biochemistry and Molecular Biology
• Bioinformatics and Computational Biology
• Drug Delivery and Pharmaceutical Sciences

Background:

• Cell-penetrating peptides (CPPs) facilitate cellular entry for therapeutic cargoes.
• Current in vivo/in vitro testing of CPPs is time-consuming and expensive.
• Existing Machine Learning (ML) models for CPPs often lack quantitative uptake data, limiting their predictive power.

Purpose of the Study:

• To create a comprehensive, open-access database of quantitative CPP uptake values and physicochemical properties.
• To develop a highly accurate ML regression model for predicting CPP uptake.
• To provide a valuable resource for accelerating CPP research and development.

Main Methods:

• Curated an open-access database (POSEIDON) with over 2,300 experimental quantitative uptake values and 1,315 peptide physicochemical properties.
• Integrated genomic features of cell lines with peptide data.
• Developed and validated an ML regression model using over 1,200 data entries.

Main Results:

• The POSEIDON database provides extensive quantitative data for CPPs.
• The developed ML regression model achieved high accuracy in predicting CPP cell line uptake.
• Performance metrics included Pearson correlation of 0.87, Spearman correlation of 0.88, and an r² score of 0.76 on an independent test set.

Conclusions:

• The POSEIDON database and ML predictor represent a significant advancement in CPP research.
• This resource offers a faster, cheaper, and more accurate alternative to traditional experimental methods.
• The freely available tools will empower researchers in designing and optimizing CPPs for therapeutic applications.