statLens Method Card
MaAsLin2: multivariable association discovery for microbiome and meta-omics
v1.26.0STATLENS VERIFIEDFinds associations between microbial features (taxa, genes, pathways) and clinical or environmental metadata. It fits a linear or generalized linear model to each feature, with built-in normalization, transformation and filtering, and random effects for repeated sampling.
Category
Regression / mixed models
Analysis task
Association, Differential analysis
Study designs
Cross-sectional, Case-control, Cohort, Longitudinal, Repeated measure, Multi-centre
Source
AI-drafted from the paper and package documentation; pending expert review
Scientific questions this method addresses
Which microbial taxa are associated with disease status after adjusting for age, sex and medication?
Which features are associated with a continuous exposure in a longitudinal cohort with repeated samples per subject?
Which pathways differ between groups in metagenomic or metatranscriptomic profiles?
Statistical specification
\log_2(y_{ij}) = \beta_0 + x_{ij}^\top \beta + b_i + \varepsilon_{ij}
Model family
Per-feature linear or generalized linear (mixed) model; default is a linear model on log-transformed, total-sum-scaled abundances
Estimand
Per-feature coefficient for each metadata variable; H₀: β_k = 0
Multiple testing
Benjamini–Hochberg FDR
Assumptions
Default pipeline: total-sum scaling, log transform, linear model; zeros are replaced before the log; Random effects capture correlation within subject or site; Features are tested independently; compositional effects are not modelled explicitly; Rare features are removed by prevalence and abundance filters
Sample size
Best suited to tens to thousands of samples
Best used when
Population-scale microbiome studies with many covariates
Longitudinal or repeated sampling, with a subject random effect
Metadata mixing continuous and categorical variables
Do not use when
The question needs absolute-abundance or explicitly compositional inference (consider ANCOM-BC or ALDEx2)
The sample size is very small relative to the number of covariates
Confirmatory work relies on the default q-value threshold of 0.25, which is lenient; set a stricter threshold
Relationships
Typed edges to other Method Cards. These are what let statLens assemble multi-step workflows rather than single methods.
supersedesMaAsLinAdds more models, normalizations and random effects
alternative_toANCOM-BCCorrects for sampling-fraction bias to target absolute abundance changes
alternative_toALDEx2Compositional approach using centred log-ratio transforms and Dirichlet sampling
validated_againstDESeq2Benchmarked against RNA-seq and microbiome methods in the MaAsLin2 paper
Outputs
Mapped to the statLens standard statistical result, plus method-specific fields.
Standard fields
feature_idestimatestandard_errorp_valueadjusted_p_valuen
Method-specific
metadatavalueN.not.zero