Annotation Input

Annotations map categorical terms to the labels present in your matrix. HiMaLAYAS filters out labels not present in the matrix and drops terms with no overlap.

Signature

Annotations(term_to_labels: dict[str, Iterable[str]], matrix: Matrix)

Parameters

Parameter	Type	Default	Description
term_to_labels	dict[str, Iterable[str]]	required	Mapping from term to labels (genes, recipes, proteins).
matrix	Matrix	required	Provides the label universe.

Common Attributes

Attribute	Type	Description
annotations.matrix_labels	set[str]	Matrix label universe used to filter incoming annotation labels.
annotations.term_to_labels	dict[str, set[str]]	Filtered term-to-label mapping retained on the object after dropping labels not present in the matrix.

Common Methods

Annotations.terms -> list[str]
Annotations.rebind(matrix: Matrix) -> Annotations

Method	Description
annotations.terms	Returns retained annotation terms after overlap filtering.
annotations.rebind(matrix)	Returns a new Annotations object aligned to a new matrix label universe (useful for zoom/subset workflows).

terms

Annotations.terms -> list[str]

Returns retained annotation terms after overlap filtering.

rebind

Annotations.rebind(matrix: Matrix) -> Annotations

Returns a new Annotations object aligned to a new matrix label universe.

Parameter	Type	Default	Description
matrix	Matrix	required	New matrix whose label universe will be used to re-filter terms.

Example (GO Biological Process)

import json
from himalayas import Annotations

with open("data/go_bp_name_to_orfs.json", "r", encoding="utf-8") as fh:
    go_bp = json.load(fh)

annotations = Annotations(go_bp, matrix)

Example (Country to Recipes)

country_to_recipes = {
    "Italy": ["r_001", "r_105", "r_214"],
    "India": ["r_002", "r_003"],
}

annotations = Annotations(country_to_recipes, matrix)

Common Errors

• Labels for term ... must be an iterable of labels if a term maps to a string.
• No annotation terms overlap matrix labels if none of the labels match the matrix.

Notes

• Terms with no overlap are dropped with a warning.
• You can pre-filter terms by size before loading.