API

We advise to import tascCODA in a python session via:

import tasccoda
ana = tasccoda.tree_ana

The workflow in tascCODA starts with reading in HTS count data and visualizing them through scCODA (sccoda.util.data_visualization). Additionally a tree structure has to be generated from hierarchical information and added to the data.

Data acquisition

Integrating data sources (dat)

Data integration works just as in scCODA. The tree structure must be added manually (see tree-structured data)

Tree data handling utilities

tasccoda.tree_utils.get_A(tree)	Calculate ancestor matrix from a toytree tree
tasccoda.tree_utils.collapse_singularities(tree)	Collapses (deletes) nodes in a toytree tree that are singularities (have only one child).
tasccoda.tree_utils.df2newick(df, levels[, …])	Converts a pandas DataFrame with hierarchical information into a newick string.

Compositional data visualization

Compositional datasets can be plotted via the methods from scCODA (sccoda.util.data_visualization).

Model setup and inference

Using the tascCODA model is easiest by generating an instance of ana.CompositionalAnalysisTree. By specifying the formula via the patsy syntax, many combinations and transformations of the covariates can be performed without redefining the covariate matrix. Also, the reference feature and the tree aggregation bias phi need to be specified in this step.

The tascCODA model

tasccoda.tree_ana.CompositionalAnalysisTree(…)	Initializer class for tascCODA models.
tasccoda.tree_agg_model_sslasso.TreeModelSSLasso(…)	Statistical model for tree-aggregated differential composition analysis (tascCODA, Ostner et al., 2021).

Result evaluation

Executing an inference method on a compositional model produces a sccoda.util.result_classes.CAResult object. This class extends the InferenceData class of arviz and supports all its diagnostic and plotting functionality. Furthermore, a function to plot the tree structure with effect labels is included.

tasccoda.tree_results.CAResult_tree(…)

Result class for tascCODA, extends the arviz framework for inference data.