Abstract <p>Fish collagen is gaining increasing attention in tissue engineering due to its exceptional bioactivity. This study aimed to isolate functional fish collagen fragments capable of microbial biosynthesis and supporting bone tissue regeneration. Collagen fragments of 150 amino acids were extracted from <i>Cyprinus carpio</i> collagen I (CcCOL1), and their bioactivity, net charge, and hydrophobicity were calculated and analyzed for correlations, these physicochemical and sequential features were using to train the machine learning model, which classified the fragments into three subgroups. Representative samples were selected from each cluster or directly from the original CcCOL1. Six out of eight variants were successfully secreted in <i>Pichia pastoris</i>, and all formed triple-helical structures, while only Var-2 and Var-3 retained self-assembly at 15&#xa0;°C. Notably, Var-2 exhibited the highest capacity to induce osteoblast differentiation. To develop scaffolds with enhanced mechanical strength, Var-2 was combined with chitin and hydroxyapatite (HAP). The resulting composite demonstrated a compressive strength of 5.77 ± 0.32&#xa0;MPa while maintaining high porosity at a chitin-HAP ratio of 2:1. Cytotoxicity assays confirmed biocompatibility, and fibroblast differentiation was comparable to Var-2 alone. In vivo rat tibia defect studies showed significant bone regeneration after 12 weeks, highlighting the potential of this fish collagen–chitin–HAP biomaterial for bone tissue engineering.</p> Graphical abstract <p>This study attempted to develop a machine learning-based method to isolate collagen fragments from <i>Cyprinus carpio</i>. This study showed the selected fragments can be secretory expressed using Pichi pastors, and the purified samples can self-assemble from triple-helix to higher-order forms, enabling its bioactivity such as inducing osteoblast differentiation and bone regeneration</p> <p></p>