VTwins

VTwins infers causative microbial features from high-dimensional metagenomic datasets by creating paired "virtual twin" samples to control confounding and enable causal inference.

Key Features:

• Confounding Elimination: Transforms the original cohort into a paired cohort of "Twin" samples that share matched taxonomic profiles but exhibit distinct phenotypes to reduce confounding effects.
• Sample Size Efficiency: Achieves high sensitivity with a reported 10-fold reduction in required sample size, validated on simulated and empirical datasets.
• Benchmark Performance: Demonstrates superior power on high-dimensional compositional datasets in benchmarks against 16 other software tools.

Scientific Applications:

• Causal inference in metagenomics: Identification of causative microbial taxa and features from shotgun or amplicon-based metagenomic studies.
• Disease-associated pathway discovery: Detection of disease-associated microbial features and pathways linked to host phenotypes.
• Microbiome–host interaction studies: Elucidation of relationships between microbial community composition and host phenotypes for hypothesis generation.

Methodology:

Transforms cohorts into paired "Twin" samples matched by taxonomic profiles but differing in phenotype, is inspired by twin studies in genetic research, and was evaluated using simulated and empirical data and benchmarked against 16 other software tools.