AI-Assisted Integration of Midgut Digestive Enzymes and Gut Microbiota in Samia ricini: A Predictive Framework for Sustainable Eri-Silk Production

Abstract

Abstract The eri silkworm, Samia ricini (Lepidoptera: Saturniidae), is an economically important polyphagous silk-producing insect whose productivity is influenced by host-plant quality, digestive physiology, larval development and associated gut microorganisms. The midgut represents a major interface between dietary nutrients and physiological utilization, while the resident microbial community may contribute to nutrient transformation and host adaptation. Although previous studies have independently investigated digestive enzymes, host-plant effects and gut microbial diversity in S. ricini, an integrated predictive framework combining biochemical, microbiome and production variables remains comparatively underdeveloped. The present study proposes an AI-assisted framework for integrating midgut digestive enzyme activity, gut microbiota and larval/cocoon traits in S. ricini. Three host-plant diets—Ricinus communis, Ailanthus altissima and Manihot esculenta—are considered within the proposed experimental framework. Digestive activities of α-amylase, protease, lipase and cellulase are proposed to be evaluated together with 16S rRNA-based bacterial community profiling. Larval growth, survival, cocoon weight, shell weight and shell percentage are incorporated as production-related variables. Statistical approaches including ANOVA, correlation analysis, diversity analysis and PERMANOVA are combined with machine-learning algorithms including Random Forest, XGBoost and Support Vector Regression. For methodological demonstration, a simulated dataset was constructed using biologically plausible ranges rather than experimental observations. The simulated analysis illustrates how integrated biochemical and microbial variables can be used to predict larval and cocoon performance and how SHAP-based explainable artificial intelligence can identify influential predictors. The framework suggests that microbiome–enzyme interactions may provide useful candidate biomarkers for future validation. However, simulated findings should not be interpreted as empirical evidence. The proposed framework provides a basis for future experimental studies aimed at microbiome-informed feeding, precision rearing and sustainable eri-silk production.