Developing a New Phylogeny-Driven Random Forest Model for Functional Metagenomics.

Area of Science:

• Microbiology
• Bioinformatics
• Computational Biology

Background:

• Metagenomics links microbial genes to functions and environmental states.
• Accurate functional classification of microbial genes is crucial for metagenomic analysis.
• Machine learning, particularly Random Forest (RF), is widely used for this classification task.

Purpose of the Study:

• To develop and evaluate a novel machine learning model, Phylogeny-RF, for enhanced functional classification of metagenomes.
• To integrate microbial phylogenetic information directly into the Random Forest algorithm.
• To improve the accuracy of metagenomic functional classification by accounting for evolutionary relatedness.

Main Methods:

• Developed the Phylogeny-Random Forest (Phylogeny-RF) model by tuning the RF algorithm with microbial phylogeny.
• Applied the Phylogeny-RF model to three real-world 16S rRNA metagenomic datasets.
• Compared Phylogeny-RF performance against traditional RF, MetaPhyl, and PhILR using metrics like AUC and Kappa.

Main Results:

• The proposed Phylogeny-RF model achieved significantly better performance than the traditional RF model (p < 0.05).
• Phylogeny-RF outperformed other phylogeny-aware benchmark methods, including MetaPhyl and PhILR.
• On soil microbiomes, Phylogeny-RF attained a highest AUC of 0.949 and Kappa of 0.891, demonstrating superior classification accuracy.

Conclusions:

• Integrating microbial phylogeny into machine learning classifiers enhances metagenomic functional classification.
• Phylogeny-RF offers a robust and accurate approach for analyzing microbial gene functions.
• The method effectively captures the impact of phylogenetic relatedness on microbial traits and functions.