Harnessing Uniform Design to Enhance AI-Driven Predictions of Physicochemical Properties of Short Peptides

Abstract

Short peptides hold significant promise in drug discovery and materials science due to their biocompatibility, multifunctionality, and ease of synthesis. However, accurately predicting their physicochemical properties, a prerequisite for application development, remains a challenge. This study presents an innovative approach integrating uniform design (UD) with artificial intelligence (AI) to enhance prediction of key physicochemical properties, including aggregation propensity (AP), hydrophilicity (logP), and isoelectric point (pI). Using UD, we generate 31 distinct peptide datasets, with a consistent amino acid occupation fraction of 5% at each position, thereby creating unbiased training data for AI models. The performance of each AI model is rigorously evaluated using various testing schemes, and optimal sample sizes are determined for accurate prediction of each property. Additionally, Shapley Additive Explanations (SHAP) analysis identifies aromaticity, logP, net charge, and pI as the primary factors affecting peptide aggregation. This work provides comprehensive datasets on the physicochemical properties of all tetrapeptides, develops robust AI-based predictive models, and elucidates the relationships between key physicochemical characteristics and self-assembly behavior. By integrating experimental design, AI modeling, and peptide domain knowledge, our approach facilitates the discovery and optimization of functional peptides, offering new opportunities for peptide-based therapeutic applications.