An improved dimensionality reduction method for meta-transcriptome indexing based diseases classification.

Area of Science:

• Microbiology
• Bioinformatics
• Computational Biology

Background:

• Bacterial 16S ribosomal RNA (rRNA) profiling is crucial for classifying microbiota-associated diseases.
• High-dimensional 16S rRNA data often exhibits sparsity and redundancy, complicating analysis.
• Traditional feature selection methods struggle with limited shared microbes and correlated abundances.

Purpose of the Study:

• To introduce a novel algorithm, Feature Merging and Selection (FMS), for effective dimensionality reduction of 16S rRNA expression data.
• To address limitations of existing methods in handling sparse and redundant microbial survey data.

Main Methods:

• The Feature Merging and Selection (FMS) algorithm integrates Linear Discriminant Analysis for feature reduction.
• FMS aims to enhance accuracy and preserve relationships between features.
• The algorithm was validated using 16S rRNA datasets from pneumonia and dental decay patients, combined with Support Vector Machine (SVM) classification.

Main Results:

• FMS demonstrated superior performance in discriminating between disease classes (pneumonia and dental caries) compared to other feature selection methods.
• The algorithm successfully reduced feature dimensions while preserving essential microbial features.
• Integration with SVM further improved classification accuracy for the tested datasets.

Conclusions:

• FMS offers a more valid and reliable approach to feature reduction in 16S rRNA data analysis.
• The algorithm improves data intelligibility by projecting high-dimensional data into a lower dimension.
• FMS enhances the classification of microbiota-associated diseases, showing promise for clinical applications.