Structural data used for morphology based prediction of neuronal functional types (Boulanger-Weill et al, 2025 - https://www.biorxiv.org/content/10.1101/2025.03.14.643363v2)

Zebrafish Hindbrain Neuron Structural Data Structural data for morphology-based prediction of neuronal functional types in the zebrafish hindbrain, as described in Boulanger-Weill et al. (2025). The dataset contains 563 neurons across three imaging modalities - photoactivation (PA), correlative light-electron microscopy (CLEM), and electron microscopy (EM) - with SWC skeleton tracings, per-cell metadata, and functional type labels. These data support an LDA classifier that predicts four functional cell types (iMI, cMI, MON, SMI) from 13 morphological features with 83.6% cross-validation accuracy. Contents File Description metadata.xlsx Cell inventory with functional labels, training flags, and cell identifiers (3 sheets: PA, CLEM, EM) paGFP.zip 47 photoactivation neurons - registered SWC skeletons, metadata, and functional dynamics (HDF5) clem_zfish1.zip 301 CLEM neurons - SWC skeletons (original and registered) and metadata em_zfish1.zip 215 EM neurons - SWC skeletons (original and registered) and metadata baselines.zip Reference prediction files and pre-computed HDF5 morphological features for regression testing custom_nblast_matrix.csv Zebrafish-trained NBLAST scoring matrix for morphological similarity verification Coordinate System CLEM and EM cells each have two SWC files: the original (*.swc, in native EM coordinates) and a registered version (*_mapped.swc, in Z-Brain atlas reference frame, Randlett et al., 2015). PA cells have a single SWC already in Z-Brain coordinates. The pipeline uses the registered skeletons. Units are microns. SWC node type labels follow standard conventions: Label Structure 1 Soma 2 Axon 3 Dendrite 4 Presynapse 5 Postsynapse Default Paths python cli.py setup --download downloads and extracts everything automatically: Platform Data path Output path macOS ~/Desktop/morph2func/morph2func_input/ ~/Desktop/morph2func/morph2func_output/classifier_pipeline/ Linux ~/morph2func/morph2func_input/ ~/morph2func/morph2func_output/classifier_pipeline/ Windows %USERPROFILE%\morph2func\morph2func_input\ %USERPROFILE%\morph2func\morph2func_output\classifier_pipeline\ Override with MORPH2FUNC_ROOT (both) or MORPH2FUNC_OUTPUT_ROOT (output only). If downloading manually, unzip into a single directory: morph2func_input/ metadata.xlsx custom_nblast_matrix.csv baselines/ paGFP/ clem_zfish1/ em_zfish1/ Usage git clone https://github.com/jboulanger91/Zebrafish_CLEM.git cd "Zebrafish_CLEM/4. Morphology_based_prediction_of_neuronal_functional_types" python cli.py env --create # Create conda environment python cli.py setup --download # Download this dataset and configure paths python cli.py run # Train classifier and predict cell types python cli.py run --help # Show all pipeline options Functional Types The four functional cell types predicted by the classifier: Abbreviation Full name iMI Motion integrator, ipsilateral cMI Motion integrator, contralateral MON Motion onset SMI Slow motion integrator Code Classification pipeline: jboulanger91/Zebrafish_CLEM / 4. Morphology_based_prediction_of_neuronal_functional_types Parent repository (registration, connectivity, modeling): jboulanger91/Zebrafish_CLEM EM connectome explorer: Neuroglancer (requires Gmail login)