<i>In silico</i> study and optimization of bacteriocin family proteins from <i>Bifidobacterium longum</i> using response surface methodology

Bacteriocins, a class of ribosomally synthesized antimicrobial peptides, have garnered significant attention in recent years for their potential applications in food preservation and therapeutic interventions. These natural compounds, produced by various bacteria, exhibit a diverse range of antimicrobial activities against closely related species and even pathogenic organisms. This study aimed to optimize the production of bacteriocin produced by <i>Bifidobacterium longum</i>. The cultivation conditions and medium composition were optimized using response surface methodology (RSM). The Plackett Burman experimental design was effective in identifying significant variables that influence bacteriocin production. The effects of main factors on bacteriocin production were further investigated by the Box-Behnken design (BBD). The predicted optimum values obtained for the maximum production of bacteriocin determined from Quadratic Machine learning methods that rely on neural networks for representation learning are known as deep learning. In the field of bioinformatics, the <i>in silico</i> study using PSIPRED software and DeepMetaPSICOV are two powerful tools used to predict protein structures and interactions. PSIPRED and DeepMetaPSICOV were utilized specifically to predict the secondary structure of bacteriocin family proteins, focusing on <i>Bifidobacterium longum</i>. The trRosetta (transform-restrained Rosetta) server software employs energy minimization algorithms to predict protein structures, while the neural network-based distance predictions help refine these structures further by considering the spatial arrangements of amino acids within the protein.