Anti-BioEn: An advanced framework for accurate bioactive agent classification based on hybrid models and graph feature encoding method

Bioactive agents are compounds that have an influence on human beings, organs, or tissues. These agents, which might be found in both natural and synthetic chemicals, are able to interact with biological systems and produce a variety of therapeutic or biological responses. In this regard, this work proposes a stacking method for categorizing five important bioactive agents: antibacterial, anti-HIV, antioxidant, antiparasitic, and antiprotozoal. This study has been designed with a graph-based feature extraction approach that successfully captures intricate interactions between molecular structures of bioactive substances. These extracted characteristics were then put into a stacking strategy, which is a strong ensemble learning technique that leverages the capabilities of several machine learning models to improve classification accuracy. By utilizing this innovative technique, the model outperformed state-of-the-art methods across all assessment criteria with more than 85% in terms of accuracy. The findings demonstrate the efficacy of the graph-based technique in conjunction with the stacking model, making it a useful tool for detecting important bioactive chemicals such as antibacterial, anti-HIV, antioxidant, antiparasitic, and antiprotozoal compounds for drug development, hence facilitating the development of novel therapeutic agents for essential health applications.