Abstract
Introduction: The traditional medicinal knowledge is one of the important resources which have not been fully explored for identification of therapeutic agents from the plant sources for diabetes mellitus. The search for reliable ethnobotanical evidence and identification of potential antidiabetic medicinal plants from the traditional medicine literature is, however, difficult because of the lack of homogeneity in traditional medicine sources, the lack of complete information about the plant's phytochemicals, and the inadequacies of the integration of computational predictions and experimental evidence.
Methods: To tackle these issues, this study suggests an integrated machine learning guided ethnobotanical discovery framework derived from publicly available medicinal plant databases, natural product repositories, scientific literature, and biological activity records. This framework consists of two novel algorithms: BOTANI-PHYTO-NET (BPN), and FUZZ-GLYCO-INSIGHT Optimization (FGIO). The main data sources are the Indian Medicinal Plants, Phytochemistry and Therapeutics database (IMPPAT 3.0) which includes 4,154 medicinal plants and 18,314 phytochemicals, Medicinal Plant Database from the Botanical Survey of India, LOTUS Natural Products Database, and bioactivity data from ChEMBL and PubChem BioAssay. Ethnobotanical knowledge extraction and identification of antidiabetic activity will be supplemented by scientific publications from PubMed and PubMed Central. BOTANI-PHYTO-NET uses natural language processing, semantic entity recognition, and network-based knowledge representation to extract medicinal plants, plant parts, traditional therapeutic uses, phytochemical constituents, and plant–compound–disease relationships. The information is then structured and represented as a DBSK within a phytomedicinal knowledge network. FUZZ-GLYCO-INSIGHT Optimization makes use of this network to obtain the estimation of antidiabetic potential and prioritize candidate plants by incorporating heterogeneous ethnobotanical, phytochemical and biological data into fuzzy inference and multi-objective optimization. The best candidates will be subjected to experimental assessment by in-vitro α-amylase and α-glucosidase inhibition assays, phytochemical characterization and cytotoxicity evaluation. The predictive framework will then be refined and validated independently with experimentally derived inhibition and toxicity measurements.
Results: The proposed approach provides an integrated computational-experimental pipeline from the traditional medicinal knowledge to the botanical network intelligence, fuzzy-based optimization and biological validation.
Conclusion: The framework is designed to enhance the systematic identification and selection of plant-derived sources of antidiabetic candidates with scientific evidence and to offer a reproducible approach for using computer-aided ethnopharmacological studies.