Harmfulness Score: A Data-Driven Framework for Ranking Environmental Risks of Microplastics

The analysis of 104,471 scientific abstracts on microplastics and nanoplastics using bibliometric tools and machine learning models produced a comprehensive mapping of thematic trends and material-specific risk associations. A composite Harmfulness Score was constructed by integrating sentiment analysis, impact descriptors, and network centrality metrics. This score ranked polystyrene (PS) and polyethylene (PE) highest in association with terms such as oxidative stress, cytotoxicity, and genotoxicity, reflecting their prominence in the literature. Reporting frequencies for key physicochemical descriptors were low-particle size (3.91%), density (0.01%), and surface area (<0.01%)-limiting their use in computational modeling and risk assessments. Thematic clustering revealed dominant topics such as environmental policy and biological impact, alongside emerging areas in microbial degradation, enzymatic transformation, and legal-policy intersections. The results highlight the need for standardized metadata practices and expanded use of analytical frameworks to enhance research reproducibility and policy relevance.