The adult hermaphrodite C. elegans nerve ring is a neuropil of 181 neurons of known identities, morphologies, contact profiles, and synaptic partners (White et al., 1986). Even for this relatively small neuropil, representations of a single feature type, such as neuronal contact profiles, constitute over 100,000 data points of multidimensional information: cell identity, region of contact, presence of synapses, etc. Analysis of this multidimensional information requires approaches that can both capture higher order patterns of organization while enabling researchers to access the underlying cell biological features resulting in these relationships. We implemented DC, a clustering algorithm that iteratively groups neurons based on the quantitative similarities of their ‘contact’ or ‘adjacency’ profiles (Brugnone et al., 2019; Moyle et al., 2021). Briefly, DC makes use of pair-wise quantifications of adjacent neuron contacts to, in a graph, move neurons with similar adjacency profiles closer together by applying a diffusion filter in a multidimensional manifold. At each iteration, the diffusion filter smooths the data across the manifold, such that local variability (or noise) in the adjacency profiles is reduced, highlighting broader, higher-order pattern similarities across neurons. As iterations proceed, individual neurons (and eventually groups of neurons) are clustered together based on how close their contact profiles are to one another in the manifold (Brugnone et al., 2019). In this way, DC uncovers hierarchical neuronal relationships in the contactome (Moyle et al., 2021).

To ensure accurate comparisons of DC across available EM datasets (Witvliet et al., 2021; White et al., 1986), we empirically determined minimum-distance adjacency thresholds (measured in pixels; Supplementary file 1) to construct adjacency profiles. Each neuron’s adjacency profile is a quantitative measure that captures the extent of contact with neighboring neurons within its spatial vicinity. By individually setting distance thresholds for each dataset, we ensured that the degree of adjacency—defined by both the presence and extent of contact—could be accurately compared across datasets generated at different times and using diverse methodologies (see also Methods and Materials), schematized in Figure 1A–C). We applied an established adjacency algorithm to quantify the extent of contact between neuron pairs by measuring the number of shared pixels within the defined distance threshold for adjacency (Brittin et al., 2021). Pixel counts were summed for each neuron across EM slices within the defined nerve ring region (see Materials and methods), resulting in an adjacency matrix representing pairwise shared pixel counts for each of the seven selected C. elegans contactome datasets (L1, 0 hours post hatch (hph); L1, 5hph; L2, 16hph; L3, 27hph; L4, 36hph; Adult 48hph (Figure 1C; See also Materials and methods). These adjacency matrices were fed into DC to reveal iterative clusters of neurons with similar adjacency profiles. To visualize and compare the results from DC, we used a graphical representation of the algorithm output called C-PHATE (Moon et al., 2019; Moyle et al., 2021), a 3-D visualization tool that builds a hierarchical, visual representation of the DC agglomeration procedure (Figure 1D–E). In C-PHATE visualizations, the DC output is mapped in 3-D space with spheres. Initially, all individual neurons in the neuropil dataset are at the periphery of the C-PHATE graph (left-hand side in schematic in Figure 1D, edges of graph in Figure 1E). Neurons are iteratively condensed together based on the similarity of their adjacency profiles (schematized in Figure 1D). In the last iteration of DC, there is a single point at the center of the C-PHATE graph which represents the entire neuropil (Figure 1E, red dot). C-PHATE representations enable visualization and comparisons of contactomes across datasets, and explorations of neuronal relationship trajectories, from individual neuron interactions to circuit-circuit bundling (Figures 1F and 2).

By Larval stage 1 (L1), 90% of neurons in the neuropil (161 neurons out of the 181 neurons) have grown into the nerve ring and adopted characteristic morphologies and positions. Although the organism grows approximately fivefold from L1 to the adult, contacts in the nerve ring are also largely established by L1 and preserved during postembryonic growth (Witvliet et al., 2021). In agreement with this, when we used DC and C-PHATE to examine contactomes from these datasets, we consistently identified four main superclusters: Stratum 1, Stratum 2, Stratum 3, and Stratum 4, using the clusters found at the highest modularity score (the iteration at which the algorithm has the highest clustering confidence) (Newman, 2006; Figure 2B–F). DC outputs for each strata across animals can also be inspected using Sankey diagrams (Supplementary file 3; Supplementary file 4; Supplementary file 5; Supplementary file 6). These diagrams detail the neuron members at each iteration of DC, allowing the user to derive quantitative comparisons of clustering events. The alignment of the neuronal morphologies of strata members reveals a persistent layered organization to the nerve ring neuropil (Figure 2G–K), and the functional identities of the neurons in each stratum suggest that there is spatial segregation of sensory information and motor outputs (Moyle et al., 2021, Supplementary file 3; Supplementary file 4; Supplementary file 5; Supplementary file 6). Our findings are consistent with previous studies on the Larval Stage 4 (L4) and adult contactomes (Moyle et al., 2021), and support that neurons establish core relationships during embryogenesis and maintain them during postembryonic growth, consistent with previous studies (Witvliet et al., 2021). Our findings also demonstrate the utility of DC and C-PHATE analyses in extracting, visualizing, and comparing the structure of the neuropil architecture across contactomes.

Because DC and C-PHATE allow for the examination of relationships at varying levels of granularity, these diagrams also facilitate the interrogation of the architectural motifs that underlie distinct neural strata. A more detailed examination of clusters reveals that while the overall strata are preserved, the underlying neuronal configurations undergo changes during postembryonic growth (Figures 2B–F–3, see:Supplementary file 3; Supplementary file 4; Supplementary file 5; Supplementary file 6). Three general features were extracted from these analyses: (1) individual neurons renegotiate their positions in the context of the identified C-PHATE clusters in different developmental contactomes, suggesting developmental changes; (2) the degree of these changes varied across the distinct strata; and (3) the degree of these changes mapped onto strata containing neurons with functions known to require either more or less neuronal plasticity, such as integrative behaviors versus more fixed reflexive behaviors, respectively. For example, Stratum 1, which contains most neurons contributing to shallow reflex circuits, controlling aversive head movements in response to noxious stimuli, displayed the fewest changes among the developmental connectomes (Figure 3B–F; Supplementary file 3). On the other hand, C. elegans exhibit tractable behaviors which can adapt due to changing environmental conditions Flavell et al., 2020. Strata 3 and 4 contain most neurons belonging to circuits associated with such learned behaviors, including chemo, mechano, and thermo sensation. This is reflected by strata 3 and 4 being the regions of the most change in neuronal relationships across postembryonic development (Figure 3G–L; Supplementary file 5; Supplementary file 6).

To examine the changes in DC/C-PHATE during postembryonic development, we made the C-PHATE plots fully interactive. This enables users to hover over and identify members of each intermediate cluster, to highlight specific cell trajectories via pseudo-coloring, and compare specific neuronal relationship dynamics across development within a multi-view window of distinct C-PHATE plots (Figure 1E-F, Figure 8—figure supplement 2; Figure 4—video 1). Because C-PHATE graphs ultimately represent cells of known identities, we reasoned that interactive mapping of the C-PHATE cluster objects to their component cellular identities and anatomies could yield greater insights on neurodevelopmental changes by linking the algorithmic abstractions of the relationships with the cell biological features and their changes across development (Figure 4).

To evaluate our hypothesis and assess the utility of C-PHATE for discovery, we examined specific regions where the distribution or ‘shape’ of super clusters changed across developmental contactomes. This approach accounts for differences in contact profiles, which directly impacts the overall clustering structure. Based on these criteria, we focused on a region displaying changes in Strata 3 and 4, and using the interactive C-PHATE graphs (Figure 4A–E), we determined the identities of neurons that changed clustering patterns across the developmental contactomes. Specifically, we focused on two interneurons, named AIML and PVQL, which we observed undergo a change in their cluster assignment from Stratum 4 (at L1) to Stratum 3 (at Larval stage 4, L4; Figure 4A and D). We pseudo-colored the trajectories of the AIML and PVQL neurons in C-PHATE to explore the changes in their C-PHATE trajectories throughout the developmental stages (Figure 4F-I, Figure 4—figure supplement 1, Supplementary file 7). Comparison of the identities of the neurons that co-cluster with AIML and PVQL suggests that the contact relationships varied across developmental stages (Figure 4F-G, Figure 4—figure supplement 1), co-cluster members can also be evaluated via Sankey diagrams, showing a switch in membership of AIM and PVQ from clusters of stratum 4 at L1-L2, pre-AVF ingrowth, to include a transitionary stratum at L3, finally to stratum 3 at L4-adult after AVF ingrowth (see Supplementary file 7).

DC/C-PHATE changes should result from changes in contact profiles. To link the observed changes in the C-PHATE graphs with the cell-biological changes in contact profiles, we generated a tool that would simultaneously enable: (1) 3D visualization of the cell-cell contact sites onto individual neuronal morphologies; (2) examination and comparisons of these contact profiles throughout development for the available contactomes; and (3) integration with DC/C-PHATE to link C-PHATE cluster objects to the 3-D morphologies of the algorithmically clustered cells. With these capabilities integrated, we could simultaneously view the contactome from two complementary perspectives – at an abstract systems level via DC/C-PHATE and at a cell biological level via 3D contact modeling – to perceive the architectural themes that underlie similar network patterns.

To create this tool, we generated 3D models of the area of physical contact between adjacent neuron pairs (Supplementary files 1 and 2; Materials and methods; Figure 5); Figure 5—figure supplement 1. Visualizing contacts from all adjacent neurons builds a multi-colored skeleton of the neuron morphology mapped onto the boundaries of this neuron (Figure 5A and C). Because the identities of the neurons are known and linked to the 3D contact models, we built text pop-ups that define the contact partners for each site (Figure 5C). Furthermore, since neuron names are consistent across the EM datasets, we can link and compare contact sites throughout development (Figure 5D). Additionally, we can analyze the representations of contact sites in the context of DC/C-PHATE clustering profiles (Figure 4F–I), 3D models of neuronal morphologies (Figure 1F), and 3D models of synaptic sites for any neuron(s) across development (Figure 7).

We used the integrated tools of DC/C-PHATE and 3D representations of the contact profiles to examine the potential cell biological changes leading to the DC/C-PHATE clustering changes observed for the AIML neuron during development. With these tools, we observed changes in the identities of the contacts made in the dorsal region of the AIML neurite (Figure 5D; Figure 5—figure supplement 2). Specifically, in the L2 stage (as compared to L1), we observed a decrease in the contacts from PVQL and an increase in contacts from the AVF neurons. This change persists to the adult stage (Figure 5D; Figure 5—figure supplement 2).

To then determine the possible source of these developmental changes in contacts, we visualized 3D models of the segmented morphologies for these neurons across L1 to adulthood (Figure 6). We find that AIM and PVQ neurons maintain similar morphologies throughout development (Figure 6C), while AVF neurons undergo substantial neurite outgrowth onto new regions of contact between AIM and PVQ (Figure 6B–D). These observations are consistent with lineaging studies which demonstrated that AVF neurons are generated from neuronal precursors (P0 and P1) at the end of the L1 stage (Sulston et al., 1983; Sun and Hobert, 2023; Poole et al., 2024; Hall and Altun, 2008; Sulston and Horvitz, 1977). By comparing the EM datasets across development, we observe that the AVF neurons grow into the nerve ring during the L2 stage and continue to grow until the Adult stage (Figure 6B–D). We also consistently observe throughout the individual datasets that the AVF neurons grow in between the AIM and PVQ neurons (Figure 6D), altering their contact profiles, which likely contributes to the observed changes in the C-PHATE graphs. While the DC/C-PHATE representations systematically cluster neurons based on relative similarities across contact profiles, and not solely by scoring changes in specific contacts within any given pair, our findings demonstrate the use of DC/C-PHATE as a discovery tool to identify cell-biological contact changes during development. (Figure 4F and G; Figure 5D; Figure 6—video 1). Consistent with this, we also observe that both AVFL and AVFR grow into the nerve ring alongside AIML, later continuing to grow around to reach AIMR, and that these relationships were reflected in the C-PHATE graphs in terms of the clustering profiles throughout development (Figure 4G; Figure 4—figure supplement 1).

We then examined if the developmental changes in contact profiles result in changes in circuitry. We examined this by layering on synaptic information. Despite dwindling AIM-PVQ contacts, AIM and PVQ neurons maintained their synaptic relationship throughout development, with synaptic sites observed primarily at the base of AIM neurons, a region of persistent contact with PVQ (Figure 7A-B). We observed that increases in contacts between AIM and AVF neurons resulted in additional en passant synapses at the new points of contact, beginning at the L2 stage and continuing to adulthood (Figure 7A–B). We also observed that AVF forms synapses with the adjacent PVQ neurons (Figure 7; Figure 7—figure supplement 1).

In summary, by integrating, representing, and comparing datasets using the new C-PHATE tools and contact profiles in NeuroSC, we identified developmental changes in the relationships of AIM, AVF, and PVQ. This case study highlights the utility of combining cell biological representations (such as morphologies, contacts, and synapses) with coarse-grained systems-level representations (like DC/C-PHATE) of vEM datasets to uncover developmental changes that could be further explored experimentally. Therefore, NeuroSC serves as a powerful platform for generating hypotheses for empirical testing, which can lead to insights into the dynamics of circuit development.

NeuroSC is built as a web-based client-server system designed to enable the sharing of anatomical connectomics data with an emphasis on facilitating the analyses of neuropil relationships across hierarchies and scales. To achieve this, we integrated tools of neuroanatomical investigation from the available C. elegans nerve ring connectomes and contactomes with a collection of 3-D modeled elements (morphologies, contacts and synapses and C-PHATE) representing different aspects of neuronal architecture and relationships (Figure 8). NeuroSC differs from other available web-based tools in this area with the integration of C-PHATE graphs that enable exploration of hierarchical organizations of stratified fascicles, the availability of new tools to examine the contactome, and the integration of these data with existing connectome and morphological datasets across developmental stages.

NeuroSC has eight key user-driven features: (1) C-PHATE, with the ability to highlight clusters containing neurons of interest (Figure 8—figure supplement 2; Figure 4—video 1), (2) interactive reconstructions of neuronal morphologies (Figure 8—figure supplement 6; Figure 6—video 2), with click-based display of cell statistics, including total volume and surface area within the defined neuropil region (see Materials and methods), (3) reconstructions of neuronal morphologies of C-PHATE cluster members with a right-click on C-PHATE clusters (Figure 4—video 1), (4) 3-D renderings of neuronal contacts to visualize the spatial distribution of contact profiles, with click-based displays of quantitative statistics, including total contact area, rank among all contacts of the primary neuron, and surface area percentages relative to both the neuron and the nerve ring (Figure 8—figure supplement 1; Figure 4—video 1), (5) 3-D representations of synaptic sites with the option to visualize subsets of those sites, and a click-based display showing the number of synapses of the selected identity within the primary neuron, including any polyadic synapse combinations involving the primary neurons (Figure 8—figure supplement 3; Figure 7—video 1) (6) the ability to perform side-by-side comparisons across development by cloning a rendered scene across time into a new window at the click of a button using the in viewer developmental stage slider (Figure 8—figure supplement 7; Figure 6—video 2). The cloning feature uses the original search query from the first viewer. ‘Contacts’ may appear or disappear when the user examines different developmental stages if the original query contacts are lost (or gained) at the specific developmental stage in the underlying data. (7) The option to pseudo color each object to highlight points of interest (Figure 8—figure supplement 7; Figure 6—video 2) and (8) each item is an individual object with the ability to be further customized by the user (Figure 8—figure supplement 7, Figure 8—figure supplement 8).

We offer seven practical considerations for users. First, NeuroSC is available on mobile platforms as a quick and convenient way to look up neuron morphologies and relationships. Second, since contact sites offer the ability to explore the surrounding neurons and the position(s) of contact between adjacent neurons, NeuroSC is designed to enable studies of adjacent neurons (e.g. phenotypes that result in site-specific ectopic synapses; neuron morphology changes that may affect specific surrounding neurons; developmental events requiring communication between neurons, etc.). Third, C-PHATE can be used to identify neurons with similar contact profiles. Because contact profiles are associated with circuit identities (Moyle et al., 2021), exploration of neuronal relationships via C-PHATE can be used to identify new relationships between specific neurons and circuits. Fourth, visualization of subsets of synaptic and contact sites allows direct comparisons to light microscopy approaches such as cell-specific labeling of synapses or GFP-Reconstitution across synaptic partners (Feinberg et al., 2008). Fifth, because the color and transparency of each 3-D model can be customized, users can further integrate NeuroSC outputs of additional atlases for gene expression, neurotransmitter and receptor expression, functional connectivity, etc. (Packer et al., 2019; Taylor et al., 2021; Wang et al., 2023; Fenyves et al., 2020; Randi et al., 2023; Maitin-Shepard et al., 2021) and directly use the NeuroSC outputs to create figures and comparisons (as done for this paper). Sixth, although synaptic sites with BWM (body wall muscles) are included in NeuroSC, the current data model limits the ability to search for these non-neuronal cells. Users can search for neurons with synapses to BWM to find this datatype. Seventh, to enable direct comparisons between our data representations and the primary EM data, the original annotations have been preserved and can be accessed by users via the sister app, CytoSHOW (CytoSHOW.org). As the data continues to be curated, the modular design of NeuroSC and its companionship with CytoSHOW enables integration of future annotations.

NeuroSC is one of several efforts centered around interpreting the C. elegans EM datasets. Other open-source tools for data exploration in C. elegans include efforts to capture neuron morphologies and synaptic information (including integration of new connectomes across larval development), to map neurotransmitter and receptor expression, and to record whole brain functional connectivity across genotypes (Witvliet et al., 2021; Altun et al., 2002; Cook et al., 2019; Fenyves et al., 2020; Randi et al., 2023; Maitin-Shepard et al., 2021; Boergens et al., 2017). NeuroSC was designed to interface and enhance these existing resources. NeuroSC focuses on comparative, biologically grounded analysis of segmented neuronal features across datasets. This includes new tools for rendering contact profiles, synchronizing 3D views, and navigating C-PHATE diagrams, all of which support multi-scale comparisons of neuronal organization. NeuroSC was designed as a series of modular, open-source tools that can be integrated onto other programs or existing resources, enhancing the existing landscape of connectomic tools. NeuroSC was inspired by tools like NemaNode and WormWiring (Witvliet et al., 2021; Cook et al., 2019), which enable 3-D visualizations of neuronal morphologies and synaptic sites with synaptic subsets restricted to pre or postsynaptic sites. In NeuroSC, we sought to generate and integrate information beyond the synaptic connectome to include local neuronal regions (contactome) and neuronal morphologies across available developmental vEM datasets. Contactomes represent features that have been largely overlooked in connectomic datasets, and which capture circuit structures not evident by inspecting solely synaptic relationships (Brittin et al., 2018; Yim et al., 2024; Cook et al., 2023). NeuroSC extends existing representations to also offer user-driven experience with choice over the visualization of specific synaptic sites, the option to search for synaptic partners, and the ability to customize the color of each synaptic representation (Figure 7). NeuroSC representations complement resource databases like WormAtlas, which hosts digitized electron micrographs and schematics of neuron morphologies with aggregated information on each neuron (Altun et al., 2002). Neuroglancer (Maitin-Shepard et al., 2021), developed by Google, is a powerful interactive tool for visualizing large-scale 3D electron microscopy (EM) and other neuroimaging datasets beyond the C. elegans community. Similar in functionality to Webknossos (Boergens et al., 2017), it provides detailed views of high-resolution EM data and enables users to explore individual datasets effectively, with a strong emphasis on synaptic connectivity. NeuroSC builds upon these capabilities by offering a complementary approach focused on comparative connectomics and broader neuronal relationships. NeuroSC enables the visualization of EM datasets alongside neuron reconstructions while allowing users to compare neuron relationships across multiple carefully curated datasets. These datasets have been standardized across time points and methodologies, making it possible to track developmental changes in neural circuits with confidence. In addition to synaptic connections, NeuroSC highlights the contactome with rendered contact patches, providing a more complete picture of spatial relationships. Users can dynamically adjust colors of rendered objects, generate scale bars for precise comparisons, and clearly visualize synapses with large geometric representations of pre- and post-synaptic structures. Importantly, neuron names and object relationships appear directly when hovering over a rendering, eliminating the need for external lookup tables. NeuroSC also supports downloading models and creating publication-ready figures, making it a valuable tool for both research and presentation. While NeuroSC does not prioritize the display of raw EM data, this information is accessible through https://NeuroSC.cytoshow.org/ (link also available on https://neurosc.net/ in the "About" section) for those who want direct EM inspection. By emphasizing comparability and standardized dataset integration, NeuroSC enables researchers to uncover insights into neural development and connectivity that might not be as easily accessible with single-dataset visualization tools. Together, existing tools and NeuroSC provide complementary ways to explore and analyze complex neural datasets, each offering unique strengths to the neuroscience community.

NeuroSC code and development was intentional in its design as an open-source resource that is modular and allows integration of additional features and data structures (Cantarelli et al., 2018). It is a hypothesis-generating tool that can be equally used by educators seeking to teach neuroanatomical principles and researchers seeking to identify changes across connectome datasets. NeuroSC could be integrated into emerging datasets, including developmental time courses of cell-specific transcriptomic data that would enable further insights on the molecular events underpinning neuronal development and function—from synaptogenic processes to the logic of neurotransmitter use (Packer et al., 2019; Taylor et al., 2021; Fenyves et al., 2020) and how it sustains functional connectivity (Randi et al., 2023). Future iterations of NeuroSC could also include positions and relationships of neurons to non-neuronal cell types, as well as the relative networks of segmented and quantified organelles within cells. NeuroSC could be used to compare new datasets from genetic variants, from animals trained under specific conditions, or from additional developmental datasets across embryogenesis. As such, the pipeline and design of NeuroSC can serve as a sandbox to examine the value of the integration of datasets in exploring representations of neuronal relationships across connectomes. Currently, the addition of new datasets must be performed in-house. In the future, standardized practices for EM generation and annotation data could facilitate the implementation of a standard importer.

NeuroSC forms part of a longer tradition that has leveraged the pioneering datasets generated for C. elegans connectomes towards exploring structure-function relationships in the nervous system. While the smaller scale of the C. elegans neuropil allowed us to rigorously vet the utility of these approaches, we suggest that these same methods would be beneficial in comparative studies in neuropils of other species, including those with less stereotypically formed connectomes. We recognize that more complex organisms possess orders of magnitude more neurons, with significantly larger neuron populations per cell class. However, our previous work has demonstrated that DC/CPHATE clustering of C. elegans neurons consistently pulls out clusters of shared neuron classes and shared functional roles (Moyle et al., 2021). Building on this foundation, we envision applying similar clustering approaches to larger connectomes, aiming to identify classes and functionally related neuronal groups in more complex nervous systems. We suggest that contact profiles, along with neuron morphologies and synaptic partners, can act as ‘fingerprints’ for individual neurons and neuron classes. These ‘fingerprints’ can be aligned across animals of the same species to create identities for neurons. Frameworks for systematic connectomics analysis in tractable model systems such as C. elegans are critical in laying a foundation for future analyses in other organisms with up to a billion-fold increase in neurons (Toga et al., 2012). We think of these collective efforts as akin to the foundational work from C. elegans in pioneering genomic analysis and annotations ahead of the Human Genome Project (Stein et al., 2001; Collins and Fink, 1995). We believe that further integration of datasets in platforms like NeuroSC would be key in determining the representations and features necessary for the interpretation and analyses of other connectomes.

Further information and requests can be directed to Daniel.colon-ramos@yale.edu.

Figures in this article have been generated with NeuroSC (Figure 5D; Figures 6 and 7, Figure 5—figure supplement 1G–I, Figure 8—figure supplement 2, Figure 7—figure supplement 1, Figure 8—figure supplement 1A–B, Figure 8, Figure 8—figure supplements 2–8; Figure 4—video 1, Figure 6—videos 1; 2, Figure 7—video 1) and CytoSHOW (Figures 1—5A and C, Figure 4—figure supplement 1, Figure 8—figure supplement 1C). Data can be visualized via the viewer at https://neurosc.net/ or by downloading glTF files from NeuroSC and using a glTF viewer to visualize them. Additionally, the data generated for NeuroSC is available in .OBJ file format at (and can be visualized from a local hard drive with CytoSHOW https://NeuroSC.cytoshow.org/). All Excel files for DC iterations and adjacency quantifications can be found in Supplementary files 3–13. Tutorials for NeuroSC are available on https://neurosc.net/ upon opening the website, within the main menu of the website (Figure 8—figure supplement 4), and in the supplementary materials (Figure 8—figure supplements 1–8; Figure 4—video 1, Figure 6—video 2, Figure 7—video 1). These tutorials generally cover the process of engaging in analysis at and across specific developmental stages by filtering the data items and adding items to viewers (Figure 8—figure supplement 6). General understanding for how to use C-PHATE to analyze neuronal relationships can be found in Figures 1 and 4, Figure 8—figure supplement 2; Figure 4—video 1, and in our previous publication (Moyle et al., 2021). For additional information on filters and in-viewer changes to the data (colors, developmental stages, downloading data), see Figure 8—figure supplement 1, Figure 8—figure supplement 3, Figure 8—figure supplement 7, Figure 8—figure supplement 8, and Figure 6—video 2, Figure 7—video 1. All code for website development is available at GitHub (https://github.com/colonramoslab/NeuroSCAN, Emerson, 2025). and for information on website architecture and data model see Figure 8—figure supplements 9–10.

Volume electron microscopy (vEM) data and segmentation of neurons and synapses were analyzed from Witvliet et al., 2021; White et al., 1986; Brittin et al., 2018; Cook et al., 2019. We analyzed available EM datasets that were transversely sectioned and segmented (Witvliet et al., 2021; Brittin et al., 2021; White et al., 1986). We deleted the CAN neurons in the L1-L3 datasets to keep these datasets consistent with the legacy datasets L4 and Adult (N2U), which do not contain CAN neurons (as in Moyle et al., 2021).

All 3-D object isosurfaces (Morphologies (Neurons), Contacts, Synapses, C-PHATE plots) were generated from segmented EM datasets using a modified version of the ImageJ 3D viewer plug-in (Schmid et al., 2010) implemented in CytoSHOW (cytoshow.org). This tool employs the marching cubes algorithm for polygon generation. All 3-D objects are first exported as wavefront (.OBJ) files, then converted to GL Transmission Format (.glTF) file format, which does not distort the resolution but compacts the file information to enable faster loading times in the web-based 3-D viewer. We intentionally avoided surface smoothing to renderings to preserve the details of the raw EM data.