Background

Bio-functional peptides (BioPeps) are short chains of amino acids with diverse biological activities, such as antimicrobial, anti-inflammatory, and anticancer properties. They are derived from various natural sources, including plants, animals, and microorganisms, and play vital roles in regulating physiological processes and maintaining health. Due to their therapeutic potential and natural origin, BioPeps are highly attractive for applications in the food, pharmaceutical, and biotechnology industries.

However, the identification and industrial-scale utilization of these peptides face significant challenges. Traditional methods of discovering and characterizing BioPeps are labor-intensive and time-consuming, relying heavily on experimental approaches that are not feasible for large-scale production. Computational models have been developed to predict BioPeps, but they often suffer from limited accuracy and generalizability. These models struggle to handle the vast diversity of peptide sequences and functions, leading to high rates of false positives and unreliable predictions.

Among the various types of BioPeps, antimicrobial peptides (AMPs) are of particular interest due to their potential as alternatives to traditional antibiotics. AMPs exhibit broad-spectrum activity against a wide range of pathogens and are less prone to resistance development, making them promising candidates for addressing the global challenge of antibiotic resistance. However, the production of AMPs in sufficient quantities remains a bottleneck, with traditional synthesis methods facing issues such as low yields, high costs, and cytotoxicity in host cells.

To address these limitations, our research focuses on developing a comprehensive deep learning model that integrates state-of-the-art Bi-directional Long-Short Term Memory (BiLSTM) networks for predicting 13 distinct categories of BioPeps. This model aims to improve prediction accuracy and reduce false-positive rates, making it more suitable for industrial applications. Additionally, we have engineered a cell-free synthesis platform to facilitate the rapid and scalable production of AMPs, overcoming the challenges associated with traditional methods.

Furthermore, to enhance the antimicrobial properties of Lactobacillus plantarum, a promising probiotic strain identified through our model, we employed directed evolution techniques. Using atmospheric and room temperature plasma (ARTP) mutagenesis and fluorescence-activated droplet sorting (FADS), we successfully generated L. plantarum mutants with significantly improved broad-spectrum antimicrobial activity and organic acid production.

This integrated approach combining advanced computational modeling, innovative synthesis methods, and experimental evolution represents a significant step forward in the field of BioPeps research. It offers a promising pathway for the discovery and industrial production of novel antimicrobial agents and other bio-functional peptides, potentially revolutionizing various sectors, from food preservation to pharmaceutical development.