STENSL

STENSL performs microbial source tracking (MST) by selecting relevant environmental sources from large-scale microbiome repositories using unsupervised environment selection and sparsity-constrained expectation-maximization to identify contributing microbial environments.

Key Features:

• Unsupervised Source Selection: Employs an unsupervised machine learning approach for environment selection to distinguish contributing from nuisance source environments.
• Sparse Identification: Integrates sparsity into the estimation process to enhance identification of true source contributions while minimizing noise from non-contributing environments.
• Expectation-Maximization Algorithm: Utilizes an expectation-maximization algorithm with sparsity constraints to infer contributing sources among large sets of potential microbial environments.
• Automated Source Exploration: Automates exploration and selection of source environments across large microbiome repositories such as the Earth Microbiome Project.

Scientific Applications:

• Ecology: Identifies environmental contributors to community composition for studies of ecosystem interactions and processes.
• Environmental microbiology: Traces sources of microbial assemblages in environmental samples to study contamination, dispersal, and habitat-specific communities.
• Public health: Attributes microbial sources relevant to pathogen tracking or exposure assessments in human-impacted environments.
• Microbial diversity and source dynamics: Enables exploration of microbial diversity and source contribution dynamics across diverse environments using large-scale repository data.

Methodology:

Performs unsupervised environment selection and sparse estimation using an expectation-maximization algorithm with sparsity constraints.