A new pipeline for structural characterization and classification of RNA-Seq microbiome data

Area of Science:

• Microbiology
• Bioinformatics
• Computational Biology

Background:

• High-throughput sequencing generates compositional data, posing challenges for traditional analysis methods.
• Analyzing interaction patterns in biological systems is crucial but complicated by relative abundance data.
• Accurate classification of samples into distinct biological groups aids in developing targeted treatments.

Purpose of the Study:

• To develop a new classification method for compositional data.
• To address the unreliability of traditional metrics with relative fractions.
• To improve the classification of biological samples into binary categories.

Main Methods:

• Developed a novel approach for classifying compositional data.
• Introduced a new metric to quantify correlation structure deviation between groups.
• Implemented dimensionality reduction for graphical representation and used simulation experiments and real 16S rRNA gene sequencing data for validation.

Main Results:

• Achieved classification accuracy of 98% or higher with synthetic data.
• Demonstrated superior performance compared to state-of-the-art methods on real microbiota datasets.
• Successfully applied the method to Operational Taxonomic Unit (OTU) count tables from 16S rRNA gene sequencing.

Conclusions:

• The proposed method offers high accuracy for classifying samples with compositional data.
• This approach effectively overcomes limitations of traditional metrics in analyzing relative abundances.
• The method shows significant promise for applications in microbiome research and clinical diagnostics.