Integrating virtual docking and neural network modeling to predict the anti-inflammatory potential of dietary polyphenols in ulcerative colitis.

While the systemic anti-inflammatory benefits of dietary polyphenols are well-established, the specific activities of structurally diverse polyphenols remain unclear. This study aimed to evaluate the regulatory capacity of plant polyphenols on inflammatory cytokines based on the affinity differences between compounds and their targets. As ulcerative colitis (UC) is characterized by intestinal inflammatory infiltration, it was selected as the entry point for this preliminary exploration. We integrated data on the intervention of 39 plant-derived bioactive components in UC from existing studies to construct a foundational dataset for the model (105 datasets in total). Based on the available data, inflammatory cytokines (TNF-α, IL-6 and IL-1β) were included as dependent variables. To enhance the assessment of intestinal inflammation, the disease activity index scores and colon length changes were also incorporated as dependent variables. Following a preliminary feasibility assessment via multiple linear regression, we developed the MPAI-SG model using a multilayer perceptron algorithm to enhance prediction accuracy. The model's performance was validated through independent studies on polyphenol-intervened UC mice. The results indicated that predicting effects based on the affinity differences between compounds and targets is feasible. The MPAI-SG model (R2 = 0.464-0.787) successfully predicted the effects of plant compounds on serum inflammatory cytokines in UC mice. Furthermore, the MPAI-SG model was utilized to predict the regulatory potential of 402 plant bioactive components on intestinal inflammation in UC mice. Leveraging the Phenol-Explorer database, 350 common foods were ranked, identifying 142 potential food ingredients for UC intervention, thereby providing a reference for dietary management of UC.