/synister_data

Synister database export (v3)

Contains

  1. All annotated synapses in synapses_v3.json Dictionary of synapse_ids to synapse attributes (skeleton_id, position, brain_region, splits)

  2. All annotated skeletons in skeletons_v3.json Dictionary of skeleton_ids to skeleton attributes (hemi_lineage_id, nt_known)

  3. All annotated hemilineages in hemi_lineages_v3.json
    Dictionary of hemilineage_ids to hemilineage attributes (nt_guess, hemi_lineage_name)

  4. All predictions for so far unknown neurotransmitters in predictions_fafb_v3_t8_p10.json Dictionary of synapse_ids to our prediction and synapse attributes (prediction, position, brain_region, splits, hemi_lineage_id, nt_known). Predictions represent class probabilities in the following order: 0: gaba, 1: acetylcholine, 2: glutamate, 3: dopamine, 4: octopamine, 5: serotonin

  5. A csv file of predicted neurotransmitters per skeleton id (CATMAID skid) after majority vote over its synapses skeleton_predictions.csv columns: skeleton_id, predicted_nt, hemilineage_id

Read files via:

ìmport json

with open('predictions_fafb_v3_t8_p10.json') as f_p:
    predictions = json.load(f_p)

To get the predicted class of a synapse run:

import numpy as np

synapse_class = np.argmax(predictions[synapse_id])

To get the predicted class of a skeleton run:

# Collect all predictions in skeleton
classes_in_skeleton = []
for synapse_id, attributes in predictions.items():
	if attributes["skeleton_id"] == skeleton_id:
		classes_in_skeleton.append(np.argmax(attributes["prediction"]))	

# Majority vote:
skeleton_class = max(set(classes_in_skeleton), key=classes_in_skeleton.count)