Discovering anti-obesity blue food compounds via combined deep learning and in silico approaches.

Blue foods, such as fish and aquatic sources, are rich in bioactive compounds with potential health benefits, including anti-obesity effects. AMPK, a key regulator of energy homeostasis and adipogenesis, plays a vital role in combating obesity by inhibiting adipocyte differentiation. This study aimed to identify natural AMPK activators from underexplored blue foods using a deep learning-based compound-protein interaction (CPI) prediction model that incorporates both local and global features of proteins and compounds. Local and global molecular characteristics were systematically represented using sequence-based encodings, molecular fingerprints, and protein descriptors to enhance the model’s ability to capture diverse biochemical relationships. The deep learning framework integrated these multimodal features through neural network modules to predict potential compound–protein interactions. In silico molecular docking was performed to refine the predicted interactions, leading to the identification of 94 top-scoring compounds with strong binding affinity toward AMPK. In vitro validation of five commercially available compounds confirmed anti-obesity activity. Notably, cyclo(Pro-Val) from Ulva pertusa, activated AMPK and modulated key adipogenic markers, suggesting its potential as a natural AMPK activator for obesity prevention and the feasibility of AI-driven screening for marine bioactives.