CCM release 2026-11
Cell-level data from six published mouse VISp datasets, across four modalities: EM connectomics, Patch-seq (excitatory and inhibitory), whole-neuron morphology and BARseq. Everything sits in one shared set of tables. The release repackages published data without reanalysis.
# read any table through its Delta log import polars as pl cells = pl.read_delta("ccm_v1.0.1_data_202611_v2/dataitem") syn = pl.scan_delta("ccm_v1.0.1_data_202611_v2/synapse") # lazy
Use a Delta reader (Polars, deltalake, Spark, DuckDB delta_scan). pandas.read_parquet(folder) ignores _delta_log/ and can return removed files as duplicate rows; for pandas, load through deltalake and convert.
Datasets
Each source is published; links go to the release, publication and portal.
What each dataset covers
Number of cells with each kind of data. Shading is the share of that dataset's cells.
EM connectivity counts cells that appear in at least one proofread-axon → single-soma connection (as presynaptic axon or postsynaptic neuron).
Cell types
Every cell-type label in the release, whether native (clustermembership) or mapped onto another taxonomy (celltoclustermapping), with cell counts in the taxonomy's own colors. Taxonomy order follows the source: Tasic cluster ids, the MET-type palette order, ABC numbering, and laminar order for EM.
EM connectomics
Proofread axons onto single-soma neurons. Embedding, type-to-type connectivity, soma positions and laminar depth for each EM volume.
Long-range projections
Two modalities measure where VISp neurons send axons: BARseq counts barcoded rolonies in targets ( neurons); WNM measures µm of reconstructed axon per CCF region ( neurons).
BARseq
WNM
Somas in CCFv3
Soma positions in the Allen CCFv3 (µm).
Taxonomies
cell-type hierarchies live in cluster. Counts are nodes per level, excluding the root.
How the tables connect
Every cell is a dataitem. Most tables join to it on its id; taxonomy tables join on (hierarchy_id, cluster).
Table dictionary
All tables, row counts from the Delta logs.