4CAC: 4-class classifier of metagenome contigs using machine learning and assembly graphs.

Area of Science:

• Microbial ecology
• Bioinformatics
• Genomics

Background:

• Microbial communities contain bacteria, archaea, viruses, plasmids, and microeukaryotes.
• Viruses, plasmids, and microeukaryotes are crucial for horizontal gene transfer and antibiotic resistance but are often overlooked due to identification challenges.
• Existing classifiers struggle with class imbalance, leading to poor identification of minor microbial classes.

Purpose of the Study:

• To develop a novel classifier, 4CAC, for simultaneous identification of viruses, plasmids, microeukaryotes, and prokaryotes in metagenome assemblies.
• To address the class imbalance issue in microbial community analysis.
• To improve the accuracy and efficiency of identifying minor microbial components.

Main Methods:

• Developed 4CAC, a classifier utilizing sequence length-adjusted XGBoost models and assembly graph information for four-way classification.
• Evaluated 4CAC on simulated and real metagenome datasets.
• Compared 4CAC's performance against existing classifiers.

Main Results:

• 4CAC significantly outperforms existing classifiers in identifying minor microbial classes, especially on short reads.
• 4CAC demonstrates an advantage on long reads, except when minor class abundance is extremely low.
• 4CAC achieves 1-2 orders of magnitude faster processing speeds compared to other methods.

Conclusions:

• 4CAC provides a substantial improvement for identifying viruses, plasmids, and microeukaryotes in metagenomic data.
• The speed and accuracy of 4CAC make it a valuable tool for microbial community analysis.
• 4CAC enhances our understanding of the roles of minor microbial components in ecological and evolutionary processes.