Enhanced Interpretable Neural Network Approach for Unified Batch Effect Mitigation and Disease Classification Using...

Area of Science:

• Microbiology
• Bioinformatics
• Computational Biology

Background:

• The oral microbiome comprises over 700 bacterial species, crucial for oral health.
• Non-biological factors introduce batch effects in microbiome sample analysis, complicating interpretation.
• Existing methods for batch effect removal and classification are often multi-step or inefficient.

Purpose of the Study:

• To develop a unified, one-step approach for simultaneous batch effect mitigation and oral microbiome disease classification.
• To evaluate the efficacy of LassoNet with batch loss in addressing batch effects and improving classification accuracy.
• To identify key oral microbiome features associated with disease outcomes.

Main Methods:

• Implementation of a novel one-step computational model integrating batch effect removal and disease classification.
• Utilized LassoNet architecture with a specific batch loss function for simultaneous processing.
• Validated the model across five oral microbiome datasets, comparing performance against baseline models.

Main Results:

• The proposed one-step method achieved an average area under the curve of 0.8 across five studies.
• Demonstrated superior performance compared to existing baseline models in oral microbiome analysis.
• Successfully identified key oral microbiome features linked to disease status through feature importance analysis.

Conclusions:

• The one-step LassoNet approach effectively addresses batch effects and classifies oral microbiome-associated diseases simultaneously.
• This method offers a more efficient and accurate alternative to traditional two-step procedures.
• The feature importance analysis provides valuable insights into microbial biomarkers for oral diseases.