Skip to content

MorphSync

MorphSync is a Python library for working with multi-layered morphological data. It provides a unified framework for managing and synchronizing different representations of morphological data such as meshes, point clouds, graphs, and tabular data.

Key Features

  • Multi-layer data management: Handle meshes, point clouds, graphs, and tables in a unified framework
  • Linking: Create and manage mappings between different data layers
  • Link-based operations: Apply masks/selections or features from one data layer to another
  • Transitivity: Easily apply transitive mappings across multiple layers
  • Lightweight: Intentionally light on algorithms, use your own tools but stop worrying about mappings

Quick Start

from morphsync import MorphSync

# Create a new morphology container
morphology = MorphSync(name=12345678)

mesh_data # object with vertices and faces
point_data # dataframe of points with optional features
graph_data # object with nodes and edges, either with optional features

# Add different layers
morphology.add_mesh("mesh", mesh_data)
morphology.add_points("synapses",
    point_data, 
    spatial_columns=["x", "y", "z"],
)
morphology.add_graph("skeleton",
    graph_data,
    spatial_columns=["x", "y", "z"],
    relation_columns=["source", "target"],
)

synapse_to_mesh_mapping # series mapping synapse indices to mesh indices
mesh_to_skeleton_mapping # series mapping mesh indices to skeleton indices

# Create mappings between layers
morphology.add_link("synapses", "mesh", mapping=synapse_to_mesh_mapping)
morphology.add_link("mesh", "skeleton", mapping=mesh_to_skeleton_mapping)

# Select all aspects of the morphology that map to the skeleton's axon labeling
# Note that transitive mapping (synapses <-> mesh <-> skeleton) works here!
mapping = morphology.query_nodes("skeleton", "compartment == 'axon'")

# Add a column to synapses which indicates the local skeleton radius
morphology.assign_from_mapping("synapses", "skeleton", ["radius"])

Core Concepts

Layers

MorphSync organizes data into layers, where each layer represents a different aspect of morphological data:

  • Table: Tabular data without a spatial component
  • Points: Point clouds such as annotations
  • Graph: Network structures like skeletons or connectivity graphs
  • Mesh: 3D surface representations with vertices and faces

Note that it's possible to have multiple layers of the same type (e.g., multiple skeleton layers).

Links define relationships and mappings between layers. They enable you to:

  • Use one layer to query or filter another layer
  • Find features from one layer based on mappings; for instance, finding the radius of skeleton nodes that synapses map to
  • Perform transitive mappings across multiple layers

Installation

pip install morphsync

or, to add to a uv managed project or environment:

uv add morphsync