mb-PHENIX: diffusion and supervised uniform manifold approximation for denoizing microbiota data

Area of Science:

• Microbiology
• Bioinformatics
• Computational Biology

Background:

• Microbiota data analysis faces challenges including technical noise, high dimensionality, and zero-inflated abundance matrices.
• These issues compromise the reliability of scientific findings and hinder the identification of differential abundance microbes.

Purpose of the Study:

• To develop an advanced algorithm for denoising microbiota data and recovering missing taxa abundances.
• To improve the accuracy of differential abundance analysis in microbiome studies.

Main Methods:

• mb-PHENIX is an open-source Python algorithm utilizing imputation via diffusion.
• It employs supervised Uniform Manifold Approximation Projection (sUMAP) for space initialization.
• This hybrid machine learning approach addresses noisy and sparse count matrices in 16S and shotgun sequencing data.

Main Results:

• mb-PHENIX effectively recovers missing taxa abundances from sparse and noisy microbiota datasets.
• The algorithm successfully denoises microbiome data, enabling the detection of differential abundance microbes.
• It outperforms traditional abundance analysis methods in revealing microbial differences among study groups.

Conclusions:

• mb-PHENIX offers a robust solution for handling common challenges in microbiota data analysis.
• The algorithm enhances the reliability and interpretability of microbiome studies.
• mb-PHENIX is accessible via GitHub with a Google Colab implementation.