Supplemental Files for Eckstein and Bates et al., Cell (2024)

<p>synister_fw_mat571_t11_synapses.feather - synapse level transmitter predictions for the FAFB dataset<br>========================================================================================</p> <p>Columns:id&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; :&nbsp;&nbsp; pre-synapse ID<br>pre&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; :&nbsp;&nbsp; neuron ID of pre-synaptic neuron<br>post&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; :&nbsp;&nbsp; neuron ID of post-synaptic neuron<br>pre_pt_position_{x, y,&nbsp; z}&nbsp; : coordinates of pre-synapse in nm<br>nts_11.*&nbsp;&nbsp;&nbsp; :&nbsp;&nbsp; predicted score for being one of the six neurotransmitter types (or "neither")</p> <p>&nbsp;</p> <p>hemibrain-v1.2-tbar-neurotransmitters.feather.bz2 - synapse level transmitter predictions for the hemibrain dataset<br>========================================================================================</p> <p>Columns:x, y, z&nbsp;&nbsp;&nbsp;&nbsp; :&nbsp;&nbsp; voxel coordinates of pre-synapse<br>conf&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; :&nbsp;&nbsp; confidence value of synapse detection<br>sv&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; :&nbsp;&nbsp; supervoxel ID of the pre-synapse<br>roi&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; :&nbsp;&nbsp; name of brain region of pre-synapse<br>roi_label&nbsp;&nbsp; :&nbsp;&nbsp; numerical label of brain region of pre-synapse<br>body&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; :&nbsp;&nbsp; body ID of the pre-synaptic neuron<br>nts_8.*&nbsp;&nbsp;&nbsp;&nbsp; :&nbsp;&nbsp; predicted score for being one of the six neurotransmitter types (or "neither")</p> <p>&nbsp;</p> <p>Neuprint_Synapses_manc_v1.ftr - synapse level transmitter predictions for the MANC dataset<br>========================================================================================</p> <p>From Takemura et al. 2023, made with the network and methods described in this repository: &nbsp;https://www.biorxiv.org/content/10.1101/2023.06.05.543757v1</p> <p>Columns:x, y, z&nbsp;&nbsp;&nbsp;&nbsp; :&nbsp;&nbsp; voxel coordinates of pre-synapse<br>conf&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; :&nbsp;&nbsp; confidence value of synapse detection<br>sv&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; :&nbsp;&nbsp; supervoxel ID of the pre-synapse<br>roi&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; :&nbsp;&nbsp; name of brain region of pre-synapse<br>roi_label&nbsp;&nbsp; :&nbsp;&nbsp; numerical label of brain region of pre-synapse<br>body&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; :&nbsp;&nbsp; body ID of the pre-synaptic neuron</p> <p>&nbsp;</p> <p>synister-master.zip - code for training and prediction with the 'synister' pipeline for neurotransmission<br>========================================================================================</p> <p>A clone of out GitHub repository at the time of paper release, found at: https://github.com/funkelab/synister, e738e22.</p> <p>Contains README with in-depth description.</p> <p>&nbsp;</p> <p>supplemental_data_1.csv - original ground truth data by identified cell type<br>========================================================================================</p> <p>A .csv file that provides each</p> <p>cell_type&nbsp;&nbsp;&nbsp;&nbsp; :&nbsp; &nbsp;cell type we used to generate our ground truth data,</p> <p>known_nt&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;the transmitter it is has been reported to express from the literature</p> <p>known_source&nbsp;&nbsp;&nbsp;&nbsp; :&nbsp; &nbsp;the study that made this report and</p> <p>evidence&nbsp;&nbsp;&nbsp;&nbsp; :&nbsp; &nbsp;the type of evidence it contained</p> <p>Note that we used this data to build our original ground truth, but have since made a wider effort to annotate known transmission for validation, reflected in Supplemental data 3 and 4, columns known_nt and known_source.</p> <p>&nbsp;</p> <p>supplemental_data_2.csv - original ground truth data by individual connectomic neuronal reconstruction<br>========================================================================================</p> <p>A .csv file that indicates each neuronal reconstruction used to generate our ground-truth data.&nbsp;Presynapses from each reconstruction were used.</p> <p>id&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;a unique identifier, which is the root ID for a flywire neuron, a skeleton ID for a FAFB-Catmaid neuron and a bodyid for a HemiBrain neuron,</p> <p>known_nt&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;the designated transmitter from the literature</p> <p>cell_type&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;its cell type and</p> <p>dataset&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;the dataset from which the neuron comes.</p> <p>FAFB-Catmaid reconstructions can be found on virtual fly brain: https://fafb.catmaid.virtualflybrain.org/. Flywire reconstructions can be found at: https://ngl.flywire.ai/. HemiBrain reconstructions can be found at: https://neuprint.janelia.org/?dataset=hemibrain. We used version 630 in this study.</p> <p>&nbsp;</p> <p>supplemental_data_3.csv -&nbsp;<br>========================================================================================</p> <p>A .csv file in which each row is a single identified HemiBrain reconstruction. Columns provide:</p> <p>bodyid&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;each neuron&rsquo;s unique identifier&nbsp;</p> <p>cell_type&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;morphological cell type</p> <p>pre&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;number of output synapses</p> <p>cropped&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;whether or not the neuronal reconstruction is cut off by the HemiBrain volume</p> <p>conf_nt&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;neuron-level transmitter prediction</p> <p>conf_nt_p&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;the confidence score for that prediction, which was calculated using the synapse-level confusion metrics and with synapse filtering</p> <p>top_nt, top_nt_p&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;neuron-level transmitter prediction with no synapse filtering for axon/dendrite</p> <p>{transmitter name}&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;the proportion of synapse-level transmitter predictions that &ldquo;voted&rdquo; for each transmitter&nbsp;</p> <p>segregation_index&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;the segregation index, a measure of neuronal polarity</p> <p>projection_score&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;the projection score, a measure of the distance between axon and dendrite</p> <p>ito_lee_hemilineage, hartenstein_hemilineage&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;the hemilineage to which this neuron belongs</p> <p>in_ground_truth&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;whether or not presynapses from this neuron were included in our ground-truth data</p> <p>&nbsp;</p> <p>supplemental_data_4.csv -&nbsp;<br>========================================================================================</p> <p>A .csv file where each row is a single identified FAFB-FlyWire reconstruction. Columns provide:</p> <p>root_id_630&nbsp; &nbsp; &nbsp;: &nbsp; each neuron&rsquo;s unique identifier from the 630 materialization of the FAFB-FlyWire dataset used in this paper</p> <p>root_id_783&nbsp; &nbsp; &nbsp;: &nbsp; each neuron&rsquo;s unique identifier from the new 783 materialization of the FAFB-FlyWire dataset used in this paper</p> <p>cell_type&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;morphological cell type</p> <p>conf_nt&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;neuron-level transmitter prediction</p> <p>conf_nt_p&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;the confidence score for that prediction, which was calculated using the synapse-level confusion metrics and with synapse filtering</p> <p>top_nt, top_nt_p&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;neuron-level transmitter prediction with no synapse filtering for axon/dendrite</p> <p>{transmitter name}&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;the proportion of synapse-level transmitter predictions that &ldquo;voted&rdquo; for each transmitter&nbsp;</p> <p>segregation_index&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;the segregation index, a measure of neuronal polarity</p> <p>projection_score&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;the projection score, a measure of the distance between axon and dendrite</p> <p>ito_lee_hemilineage, hartenstein_hemilineage&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;the hemilineage to which this neuron belongs</p> <p>side, morphology group, flow, cell class, cell sub class &nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;other metadata from Schlegel et al. 2023, https://zenodo.org/records/8077335</p> <p>in_ground_truth&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;whether or not presynapses from this neuron were included in our ground-truth data</p> <p>notes&nbsp; &nbsp; &nbsp;: &nbsp; extra notes on the cell type designation, including alternate names</p> <p>pos_x, pos_y, pos_z&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;because the root id for neurons is changing as neurons are edited in an active connectome project we also supply the position of a point in the neuron to help identify it in FAFB-FlyWire voxel space</p> <p>nucleus_id&nbsp; &nbsp; &nbsp;: &nbsp; an ID for the nucleus segmentation</p> <p>&nbsp;</p> <p>supplemental_data_5.zip -&nbsp;<br>========================================================================================</p> <p>A .zip archive containing .png files depicting each of the 183 brain hemilineages we have used from the FAFB-FlyWire dataset. Neurons in each hemilineage are colored by their neuron-level transmitter predictions, hemilineage names given in the file name. Hemilineage labels for the FAFB-FlyWire dataset are fully reported in Schlegel et al. 2023, https://zenodo.org/records/8077335.</p> <p>&nbsp;</p> <p>supplemental_data_6.csv -&nbsp;<br>========================================================================================</p> <p>A .csv containing summary results for our 183 central brain secondary hemilineages. Columns provide:</p> <p>ito_lee_hemilineage&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;each hemilineage&rsquo;s name&nbsp;</p> <p>hartenstein_hemilineage&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;and lineage-associated tract</p> <p>left, right, center&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;the number of neurons on the left and right hemispheres (left, right, center),</p> <p>lineage_type&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;the lineage type i.e. Type I or II</p> <p>{transmitter}&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;the number of neurons predicted for each transmitter</p> <p>majority_nt&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;and the majority vote of the hemilineage.</p> <p>Note, the same information is given for HemiBrain hemilineages, with &lsquo;hemibrain &rsquo; added to the column name.</p> <p>&nbsp;</p> <p>supplemental_data_7.csv -&nbsp;<br>========================================================================================</p> <p>A .csv in which each row is a described cell type of the fly brain. Cell type annotations were drawn from Schlegel et al. 2023, neurons with no cell type or hemibrain type annotated are not included. Columns provide:</p> <p>cell_type&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;each cell type&rsquo;s name</p> <p>cell_type_nt&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;predicted transmitter for the cell type (cell type nt)</p> <p>cell_type_confidence&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;the cell type level transmitter confidence score</p> <p>{transmitter}_{dataset}_confidence&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;the sum of neuronlevel transmitter predictions across all FAFB-FlyWire reconstructions for each transmitter</p> <p>{transmitter}_{dataset }_n&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;the number of neurons with that confidence,</p> <p>known_nt&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;the transmitters reported to be used by the cell type in the literature</p> <p>known_nt_source&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;the citations for that information&nbsp;</p> <p>ito_lee_hemilineage, hartenstein_hemilineage, morphology_group, flow, cell_class, cell_sub_class&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;morphological annotations from Schlegel et al. 2023</p> <p>cell_type_nt&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;the cell type level transmitter is determined by the highest {transmitter} cell type {dataset} confidence</p> <p>cell_type_nt_conf&nbsp; &nbsp; &nbsp;:&nbsp; &nbsp;The cell type level transmitter confidence score (cell type nt) was calculated using our prediction confusion matrices.</p> <p>Results for both FAFB-FlyWire (&lsquo; fafb&rsquo;) and HemiBrain (&lsquo;hemibrain&rsquo;) datasets are given.</p> <p>&nbsp;</p> <p>supplemental_data_8.tar.gz<br>========================================================================================</p> <p>Contains PNG files of 2D images cropped around synapses, together with their<br>counterfactual translations into different neurotransmitter types and<br>highlights of the most important areas for classification.There is one PNG file for each pair of neurotransmitters, named<br>`{source}_{target}.png`. E.g., the file `ach_dop.png` contains translations of<br>real cholinergic synapses into counterfactual dopaminergic synapses.Each PNG shows images in six columns, which are:1: original raw + classifier scores<br>2: counterfactual raw + classifier scores<br>3: counterfactual raw, with original raw in mask + classifier scores<br>4: mask, shown on original + relative classifier change<br>5: mask, shown on counterfactual<br>6: difference in mask area</p> <p>Paper: https://www.cell.com/cell/fulltext/S0092-8674(24)00307-6</p>