Loader Group

The groupby() of subtomogram loaders returns a LoaderGroup object. An LoaderGroup object is very similar to those returned by groupby() methods of polars.DataFrame, Molecules or pandas.DataFrame.

from acryo import SubtomogramLoader

loader = SubtomogramLoader(image, molecules)
for cluster, ldr in loader.groupby("cluster_id"):
    assert (out.molecules.features["cluster_id"] == cluster).all()

LoaderGroup has many methods of the same name as those in SubtomogramLoader.

Group-wise Averaging

LoaderGroup supports all the averaging methods.

• average()

• average_split()

In LoaderGroup version, result is returned as a dict of group key and the averages.

Group-wise Alignment

LoaderGroup also supports all the alignment methods

• align()

• align_no_template()

• align_multi_templates()

In LoaderGroup version, result is returned as an updated LoaderGroup.

If you want to collect aligned Molecules objects, following codes are essentially equivalent.

# call align() for each loader
aligned = []
for cluster, ldr in loader.groupby("cluster_id"):
    out = ldr.align(template)
    aligned.append(out.molecules)

# call align() of the LoaderGroup object.
aligned = []
for cluster, ldr in loader.groupby("cluster_id").align(template):
    aligned.append(out.molecules)

Since each group does not necessarily composed of the same molecules, you can use a mapping of templates for alignment functions.

templates = {
    0: template0,
    1: template1,
    2: template2,
}
aligned = loader.groupby("cluster_id").align_multi_templates(templates)