Machine learning reveals microbial interactions driving plastic degradation across plastisphere environments.

Area of Science:

• Environmental microbiology
• Microbial ecology
• Biotechnology

Background:

• Microplastic pollution is a growing global concern, leading to the formation of distinct microbial communities on plastic surfaces, known as the plastisphere.
• The composition and function of these plastisphere communities vary significantly across different environmental settings.
• Understanding the interactions within plastisphere communities, particularly between plastic-degrading bacteria (PDBs) and non-plastic-degrading bacteria (NDBs), is crucial for plastic biodegradation research.

Purpose of the Study:

• To investigate the microbial diversity and community structure of the plastisphere across ocean, surface water, and wastewater habitats.
• To explore the interactions between potential plastic-degrading bacteria (PDBs) and non-plastic-degrading bacteria (NDBs) within these diverse environments.
• To evaluate the effectiveness of machine learning (ML) approaches in identifying key taxa and ecological interactions in plastisphere communities.

Main Methods:

• 16S rRNA gene sequencing was employed to profile microbial communities.
• Machine learning (ML) algorithms, including Random Forest modeling, were utilized for data analysis.
• Network analyses, such as Pearson's correlation, were performed to uncover co-occurrence patterns and potential interactions between bacterial taxa.

Main Results:

• Wastewater plastispheres exhibited the highest microbial diversity and compositional evenness, attributed to complex nutrient loads and pollutant inputs.
• Habitat-specific potential PDBs were identified, with distinct genera found in wastewater (e.g., *Pseudomonas*), ocean (e.g., *Flavobacterium*), and surface waters (e.g., *Psychrobacter*).
• Consistent co-occurrence patterns between potential PDBs and various NDB taxa were observed, suggesting facilitative interactions like nutrient exchange and biofilm support.

Conclusions:

• Plastisphere communities are ecologically complex, with significant variations across different aquatic environments.
• Wastewater environments represent hotspots for diverse plastisphere microbial communities with potential for plastic biodegradation.
• Machine learning tools show promise for identifying microbial interactions but require further development for functional validation and integration of environmental metadata for comprehensive ecological insights.