What does this strategy enable?
Model paired microbial and MS metabolomics measurements to prioritize conditional co-occurrence relationships between taxa and molecular features.
Transferable workflow
Workflow
- 1
Match the two measurement tables
Align microbial and metabolite features using shared sample identifiers; inspect filtering, sparsity and study design.
- 2
Fit and evaluate the conditional model
Use mmvec with appropriate tuning and held-out evaluation; inspect training diagnostics.
- 3
Prioritize and validate associations
Inspect conditional rankings and ordinations, then design independent tests for biological hypotheses.
Connected resources
Tools used in this strategy
Strategy profile
Related objectives
Phenotype association
Components
Metadata analysis, Statistical analysis
Scope and considerations
Designed for
- Associations across paired microbiome and MS metabolomics samples.
Important requirements
- Matched microbial feature and metabolite abundance tables with sufficient shared samples.
- Recorded preprocessing and a validation design appropriate to the study.
Considerations
- Conditional co-occurrence does not establish metabolite production, consumption or causation.
- A low conditional score is not automatically a negative correlation.
- Compare against suitable compositional-data baselines; universal superiority over linear methods is not established. See the 2021 discussion (10.1038/s41592-020-01006-1) and reply (10.1038/s41592-020-01007-0).
- Small studies and sparse data require careful tuning and uncertainty assessment.
Reusable resources
✓ Code available✓ Published method available
Publication
Morton JT, Aksenov AA, Nothias LF, et al. Nature Methods 16, 1306-1314 (2019).