Uniform design-embedded predictions of (tetra-)peptide physicochemical properties

Area of Science:

• * Peptide science
• * Computational chemistry
• * Drug discovery

Background:

• * Short peptides offer significant potential in medicine and materials science due to their favorable characteristics.
• * Predicting peptide physicochemical properties is crucial but challenging due to the vast number of possible sequences.
• * Accurate property prediction is essential for developing peptide-based therapeutics and materials.

Purpose of the Study:

• * To develop an efficient and accurate method for predicting physicochemical properties of short peptides.
• * To create comprehensive datasets for tetrapeptide physicochemical properties.
• * To elucidate structure-property relationships in short peptides.

Main Methods:

• * Integration of Uniform Design (UD) for strategic data sampling across the entire sequence space.
• * Application of Artificial Intelligence (AI) models trained on UD-generated datasets.
• * Analysis using Shapley Additive Explanations (SHAP) to understand attribute contributions.

Main Results:

• * Generation of 31 distinct, unbiased tetrapeptide datasets using UD.
• * Development of robust AI models for predicting aggregation propensity (AP), hydrophilicity (logP), and isoelectric point (pI).
• * Quantitative elucidation of relationships between peptide attributes and self-assembly behavior.

Conclusions:

• * The UD-AI approach effectively predicts key physicochemical properties for a large number of peptide sequences.
• * This integrated methodology facilitates the discovery and optimization of functional peptides.
• * The findings open new avenues for peptide-based therapeutic development.