A systematic longitudinal study of microbiome: integrating temporal-spatial dimensions with causal and deep learning...

Area of Science:

• Microbiome research
• Computational biology
• Systems biology

Background:

• Longitudinal microbiome data offer insights into host-microbe dynamics during disease.
• Challenges include missing values, sparsity, and interpretability, hindering biomarker discovery and disease modeling.

Purpose of the Study:

• Introduce SysLM, a deep learning framework for systematic analysis of longitudinal microbiome data.
• Address missing values and enhance biomarker discovery for improved disease modeling.

Main Methods:

• SysLM-I module: missing-value imputation using metadata, feature enhancement, Temporal Convolutional Network, and Bi-directional Long Short-Term Memory.
• SysLM-C module: integrates deep learning with causal inference for classification and biomarker screening (differential, network, core, dynamic, disease-specific, shared).

Main Results:

• SysLM shows superior performance in imputation, classification, and biomarker discovery across multiple datasets.
• Identifies novel microbial mechanisms in ulcerative colitis, demonstrating potential for precision medicine.

Conclusions:

• SysLM provides a robust framework for analyzing complex longitudinal microbiome data.
• Integrating deep learning with causal modeling advances microbiome-based disease research and therapeutic strategies.