Machine Learning in Environmental Metagenomic Approaches: A Bibliometric and Thematic Analysis

Abstract

Machine learning is increasingly integrated with metagenomic analysis, yet the development, conceptual organization, environmental applications, and recent thematic directions of this research area require systematic characterization. This study examined English-language article and review records indexed in the Web of Science Core Collection from 2016 to 2025 using a predefined search strategy. In this study, bibliometric and science-mapping analyses were conducted to examine annual scientific production, source productivity, keyword co-occurrence, thematic structure, trend topics, and geographic and collaboration patterns. The analysis covered 286 documents published across 139 sources, with an annual growth rate of 54.79%, and publication output reached 102 documents in 2025. Metagenomics and machine learning formed major conceptual foundations, with the thematic structure linking computational methods to microbial ecology, diversity, soil, and bacterial communities. Keyword co-occurrence further distinguished interconnected groupings centered on machine learning–metagenomics–bioinformatics, soil–rhizosphere and nutrient-related research, bioremediation–biodegradation, and resistome–wastewater surveillance. Recent trend topics included artificial intelligence, prediction, and degradation. When considered together with the thematic-map position of degradation–bioremediation–artificial intelligence, these patterns support a cautious interpretation that this research direction is recently gaining visibility rather than representing a definitively emerging or dominant research front. Scientific production was geographically concentrated, particularly in China and the United States, while international collaboration was also evident. Overall, the field shows increasing methodological diversity, with reference-based approaches being complemented by predictive and explainable methods.