An Exponential Scale Mixture Model for Metatranscriptomics Data with Application to Inflammatory Bowel Disease.

Area of Science:

• Microbiology
• Bioinformatics
• Genomics

Background:

• Metatranscriptomic (MTX) sequencing profiles microbial gene expression, linking genetic potential to functional activity.
• Standard MTX pipelines often report normalized abundances, hindering count-based RNA-seq methods.
• Existing Gaussian models for MTX data involve transformations and assumptions unsuitable for sparse data.

Purpose of the Study:

• To propose a novel statistical framework for differential gene expression analysis in MTX data.
• To address limitations of current methods, including normalization, sparsity, and genomic potential.
• To identify novel transcriptomic patterns associated with dysbiosis in inflammatory bowel disease (IBD).

Main Methods:

• Developed a new modeling framework using a scale mixture of exponential distributions for MTX differential expression analysis.
• Incorporated DNA abundance to adjust for genomic potential.
• Accounted for subject-specific random effects, treated zeros as left-censored, and used a mixture prior for sparsity.
• Applied the model to the IBDMDB multi-omics cohort.

Main Results:

• Differential expression results varied significantly across different models, including Gaussian approaches.
• The proposed model identified a distinct subset of candidate genes missed by existing Gaussian methods.
• Effect directions for dysbiosis were consistent with Gaussian models, but effect sizes were larger in absolute value.

Conclusions:

• The new modeling framework offers improved analysis of MTX data, particularly for sparse datasets.
• This approach can reveal novel insights into transcriptomic dysbiosis in IBD.
• The findings highlight the importance of appropriate statistical modeling for MTX data interpretation.