def _ilr_with_tree(X, tree):
# Import ilr_transform from gneiss
from gneiss.composition import ilr_transform
# Check tree type
tree_type = _infer_tree_type(tree)
# Check leaves
components = set(X.columns)
leaves_in_tree = _get_leaves(tree=tree, tree_type=tree_type)
assert components <= leaves_in_tree, "Not all components (X.columns) are represented in tree"
# Prune tree
if components < leaves_in_tree:
tree = tree.copy()
n_leaves_before_pruning = len(leaves_in_tree)
tree = _prune_tree(tree=tree,
tree_type=tree_type,
leaves=components)
n_leaves_after_pruning = len(
_get_leaves(tree=tree, tree_type=tree_type))
n_pruned = n_leaves_before_pruning - n_leaves_after_pruning
if verbose:
print("Pruned {} attributes to match components (X.columns)".
format(n_pruned),
file=sys.stderr)
# Polytomy
if check_polytomy:
_check_polytomy(tree=tree, tree_type=tree_type)
# ETE
if tree_type == "ete":
tree = _ete_to_skbio(tree=tree)
return ilr_transform(table=X, tree=tree)
def ilr_phylogenetic(
table: pd.DataFrame,
tree: skbio.TreeNode,
pseudocount: float = 0.5) -> (pd.DataFrame, skbio.TreeNode):
t = tree.copy()
t.bifurcate()
t = rename_internal_nodes(t)
return ilr_transform(add_pseudocount(table, pseudocount), t), t
def ilr_hierarchical(table: pd.DataFrame,
tree: skbio.TreeNode,
pseudocount: float = 0.5) -> pd.DataFrame:
return ilr_transform(add_pseudocount(table, pseudocount), tree)