Microbiome profile data representation (MPDR)

mllabiome works with taxonomic abundance profiles in tab-separated format. MPDR is the data representation used throughout the framework to describe which features are extracted from the raw profiles and in which form they are presented to the model.

An MPDR is defined by three orthogonal dimensions:

• Taxonomic levels — which clade ranks are included as features
• Transformation type — how raw abundance values are transformed before training
• Projection method — an optional dimensionality reduction step applied after transformation

Each dimension is configurable independently. The subpages in this section describe the available options and the corresponding predefined configuration sets.

For the expected input file format, see Data preparation.