DEBIAS-M

DEBIAS-M corrects processing bias in microbiome data by learning interpretable bias-correction factors for each microbe in each batch that simultaneously minimize batch effects and maximize cross-study associations with phenotypes.

Key Features:

• Interpretable bias-correction factors: Infers multiplicative correction factors for each microbe in each batch that are stable and strongly associated with specific experimental protocols.
• Domain adaptation: Facilitates cross-study integration by correcting for processing biases without requiring strong parametric assumptions.
• Phenotype-guided correction: Maximizes cross-study associations with phenotypes while minimizing batch effects, without requiring an outcome variable that risks overfitting.
• Preserves sparsity structure: Avoids introducing non-biological values to features that were initially mostly zeros, unlike many existing transformations.

Scientific Applications:

• HIV classification: Improves prediction of HIV status from gut microbiome data across studies.
• Colorectal cancer classification: Enhances cross-study generalization of colorectal cancer prediction models from gut microbiome data.
• Cervical neoplasia prediction: Improves prediction of cervical neoplasia from cervical microbiome data across different studies.
• Cross-study biomarker discovery: Enables identification of reproducible biological signals and interpretable microbiome-based prediction models that generalize across independent studies.

Methodology:

DEBIAS-M learns bias-correction factors for each microbe in each batch through an optimization framework that simultaneously minimizes batch effects and maximizes cross-study associations with phenotypes.