kLDM

kLDM infers multiple metagenomic association networks that vary with environmental factors to identify condition-specific microbe-microbe and environmental factor–microbe associations.

Key Features:

• Dynamic Network Inference: kLDM estimates multiple association networks corresponding to specific environmental conditions, capturing fluctuations in microbial interactions with changing EFs.
• Simultaneous Association Estimation: The model simultaneously infers both microbe-microbe and environmental factor-microbe associations within each condition-specific network.
• Versatility Across Datasets: kLDM has been validated on synthetic data, colorectal cancer samples, TARA Oceans, and American Gut project datasets.
• Comparison with Traditional Methods: kLDM outperforms methods such as Spearman’s rank correlation coefficient (SCC) by accounting for environmental variability and condition-specific sample separation.

Scientific Applications:

• Colorectal Cancer Research: Analysis of fecal samples from cancer patients revealed fewer overall microbial associations but stronger connections between specific bacteria associated with colorectal cancer (CRC), suggesting gut microbe translocation.
• Marine Ecosystem Studies: Identification of environmental-factor-dependent associations within marine eukaryotic communities, informing ecological dynamics in oceanic systems.
• Gut Microbial Heterogeneity: Detection of significant microbial heterogeneity in studies of irritable bowel disease (IBD) patients, providing insights relevant to understanding and treating gut-related disorders.

Methodology:

kLDM implements a k-Lognormal-Dirichlet-Multinomial statistical model that integrates environmental factor data with metagenomic abundance profiles to construct multiple condition-specific association networks and estimate microbe-microbe and environmental factor-microbe associations.