Standard TaxSEA works at the group level. You run a differential abundance analysis, get a fold change for each taxon, and ask whether biologically related taxa shift together. That’s powerful for characterising what differs between groups.
But sometimes the more interesting question is about individuals:
Group-level enrichment can’t answer these questions because it collapses all individuals into a single summary statistic. ssTaxSEA extends TaxSEA to produce an enrichment score for every sample, for every taxon set.
The core idea is borrowed from single-sample Gene Set Enrichment Analysis (ssGSEA), adapted for microbiome count data.
The approach has four steps:
CLR transform. Raw counts are transformed using the centered log-ratio, which accounts for the compositional nature of microbiome data. Zeros are replaced with a small pseudocount (0.5) before transformation.
Z-score each taxon across the cohort. After CLR transformation, each taxon is z-scored across all samples. This is the critical step. It means that values no longer reflect within-sample abundance, but instead reflect how much higher or lower a taxon is in a given sample relative to the cohort mean. Without this step, you would only be asking “are these taxa the most abundant in this person?” rather than “does this person have unusually high levels of these taxa compared to everyone else?”.
Rank and score. For each sample, the z-scores are ranked and an ssGSEA-style weighted running-sum enrichment score is computed for each taxon set.
KS test. A Kolmogorov-Smirnov p-value is computed per sample per set, testing whether the set members are shifted relative to all taxa in that sample.
The output is a matrix of enrichment scores (samples x taxon sets) and a corresponding matrix of p-values.
This is worth emphasising because it’s the key difference between a naive per-sample approach and one that actually works.
Consider oral bacteria in an IBD cohort. At the group level, these taxa are consistently elevated in IBD relative to healthy controls, and TaxSEA picks this up easily from fold changes. But in any individual sample, oral bacteria are unlikely to be the most abundant taxa. They might sit at moderate abundance in IBD samples and low abundance in controls.
If you simply rank each sample’s taxa by their own abundance and run enrichment, oral bacteria won’t stand out in either group because both groups have the same dominant gut commensals. The within-sample ranking buries the between-group signal.
Z-scoring across the cohort solves this. A taxon gets a high z-score in a sample only if that sample has more of it than average. So in IBD samples, oral bacteria will have positive z-scores (elevated relative to the cohort), and in healthy samples, they will have negative z-scores. The enrichment test then picks up exactly the pattern you expect.
| TaxSEA | ssTaxSEA | |
|---|---|---|
| Input | Named vector of fold changes or correlations | Count matrix (taxa x samples) or TreeSummarizedExperiment |
| Output | One enrichment result per taxon set | One enrichment score per sample per taxon set |
| Question | What trait-based shifts characterise the difference between groups? | Which individual samples show a given enrichment pattern? |
| Use case | Interpreting DA results, contextualising group differences | Patient stratification, individual-level phenotyping, heterogeneity analysis |
In practice, you might use both: run TaxSEA first to identify which taxon sets are enriched at the group level, then use ssTaxSEA to see how those enrichments distribute across individuals.
The input should be a matrix with taxa as rows and samples as columns. Row names must be taxon names (species or genus) in the same format as standard TaxSEA.
ssTaxSEA accepts TSE objects directly. It will extract the count assay and use rownames() as taxon identifiers.
As with standard TaxSEA, you can provide your own taxon sets. When using a custom database, names in the sets should match your row names directly (no NCBI ID conversion is performed).
my_sets <- list(
oral_taxa = c("Fusobacterium_nucleatum", "Streptococcus_mutans",
"Porphyromonas_gingivalis", "Prevotella_intermedia",
"Treponema_denticola", "Aggregatibacter_actinomycetemcomitans"),
scfa_producers = c("Faecalibacterium_prausnitzii", "Roseburia_intestinalis",
"Eubacterium_rectale", "Coprococcus_catus",
"Butyricicoccus_pullicaecorum", "Anaerostipes_caccae")
)
res <- ssTaxSEA(counts, custom_db = my_sets, min_set_size = 3)Scores. A positive enrichment score for a given sample and taxon set means that the taxa in that set tend to have higher abundance in that sample relative to the rest of the cohort. A negative score means they tend to be lower. The magnitude reflects how strong and coordinated the shift is.
P-values. The KS test p-value tells you whether the distribution of z-scores for set members differs from the background in that sample. Note that these are unadjusted p-values. Since you are testing many sets across many samples, apply your own multiple testing correction as appropriate for your analysis.
Cohort composition matters. Because z-scoring is relative to the cohort mean, the composition of your cohort affects the scores. If your cohort is 90% disease samples, the “average” will look disease-like and scores will be less extreme. A balanced cohort (or at least a well-defined reference group) generally gives more interpretable results.
Sample size. Z-scoring requires a reasonable number of samples to estimate the mean and standard deviation for each taxon. With very few samples (fewer than ~10), estimates will be noisy.
BugSigDB is excluded. ssTaxSEA uses the built-in TaxSEA database (GMRepoV2, MiMeDB, gutMGene, mBodyMap, BacDive) but does not load BugSigDB. This keeps the function fast and avoids the need for the bugsigdbr dependency.
Downstream analysis. The score matrix is designed to slot into standard statistical workflows. You can use it for: