Connectomic reconstruction from hippocampal CA3 reveals spatially graded mossy fiber inputs and selective feedforward inhibition to pyramidal cells
The repository is for the following preprint.
(https://www.biorxiv.org/content/10.1101/2025.07.09.663979v1)
The EM image data and segmentation can be accessed at (https://pyr.ai)
A list of all identified cell types can be found at (https://codex.flywire.ai/research/mouse_ca3_explorer)
The codes were written using Python version 3.10.14.
The python module versions used for the analysis are in requirements.txt file.
Create a virtual environment. Then install all dependencies using:
pip install -r requirements.txt
The repository is not set up for pip installation.
All notebook codes in this repo do not require GPUs. GPUs are required for vesicle segmentation, which was accomplished by using the software called Cellpose.
Training on one NVIDIA GTX 1050 Ti GPU took about 2 hours
The EM and segmentation datasets are too large for conventional downloads. They can be accessed via cloudvolume, a tool that faciliates a numpy-like interface to the data for subvolume reading. Data is periodically updated and versioned. If you want to reproduce the results of the paper, use the materialization version 195.
-
Access the 195 version segmentation volume using:
from cloudvolume import CloudVolume as cv
vol =cv('gs://zheng_mouse_hippocampus_production/v2/seg_m195',parallel=True, progress=False, use_https=True) -
Access the mesh for a specific segment ID
mesh = vol.mesh.get(648518346446845973) -
Access skeleton
skeleton = vol.skeleton.get(648518346446845973)
Please reach out to the Seung Lab if you need a copy of entire volumetric dataset (e.g. to train machine learning models). Triangulated meshes are available for all segments in the datasets (proofread and unproofread) and can be download on a per-segment basis.
Tutorial
CAVE is the software infrastructure used to host the CA3 dataset and the primary access point for programmatic queries. Through CAVE synaptic connections and annotations can be queried.
To see a demo of MF bouton extraction, run the notebook script Demo_MF_bouton_extraction.ipynb
For bouton extraction, you need both presynaptic neuron ID (i.e. MF) and postsynaptic neuron ID (i.e. Pyr cell).
You can change the cells by changing the values in pyr_ids and mf_ids variables.
The bouton extraction result will be saved as an .html file: 
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
