SenNet Portal: Build, Optimization and Usage

Abstract

Cellular senescence is a hallmark of aging and a driver of functional decline across tissues, yet its heterogeneity and context dependence have limited systematic study. The Common Fund’s Cellular Senescence Network (SenNet) Program addresses this challenge by generating multimodal, multi-tissue datasets that profile senescent cells across the human lifespan and complementary mouse models. The SenNet Data Portal (https://data.sennetconsortium.org) serves as the public gateway to these resources, providing open access to harmonized single-cell, spatial, imaging, transcriptomic, and proteomic data; senescence biomarker catalogs; and standardized protocols that can be used to comprehensively identify and characterize senescent cells in mouse and human tissue. As of January 2026, the portal hosts 1,753 publicly available human and mouse datasets across 15 organs using 6 general assay types. Experts from 13 Tissue Mapping Centers (TMCs) and 12 Technology Development and Application (TDAs) components contribute tissue data, analyze data, identify senescent biomarkers, and agree on panels for cross-tissue antibody harmonization. They also register human tissue data into the Human Reference Atlas (HRA) and develop user interfaces for the multiscale and multimodal exploration of this data. Built on a scalable hybrid cloud microservices architecture by the Consortium Organization and Data Coordinating Center (CODCC), the Portal enables data submission, management, integrated analysis, spatial context mapping, and cross-species senescence mapping critical for aging research. This paper presents user needs, the Portal’s architecture, data processing workflows, and senescence-focused analytical tools. The paper also presents usage scenarios illustrating applications in biomarker discovery, quality benchmarking, hypothesis generation, spatial analysis, cost-efficient profiling, and cell distance distribution analysis. Current limitations and planned extensions—including expanded spatial-omics releases and improved tools for senotype characterization—are discussed. SenNet protocols, code, and user interfaces are freely available on https://docs.sennetconsortium.org/apis.

Inaugurated in 2021, the Common Fund’s Cellular Senescence Network (SenNet) Program was established to comprehensively identify and characterize senescent cell identities and abundances across the body, across various states of human health, and across the lifespan.¹ Resources and results currently being produced by SenNet include publicly accessible atlases of senescent cells and descriptions of their shared and distinct molecular identities, see https://sennetconsortium.org for resources, data, and publication menus. Data collected from human and mouse model organism tissues were used to identify and characterize senescent cells.² Novel tools and technologies were designed, implemented, and validated in close collaboration with experts from while expanding methods and technologies developed by the Common Fund’s Human Biomolecular Atlas Program^3,4 and Single Cell Analysis Program.⁵

There are three SenNet portal websites: The SenNet Consortium Website (https://sennetconsortium.org) introduces the SenNet consortium and provides public access to resources, news, working groups, and consortium member services. The SenNet Data Portal (https://data.sennetconsortium.org) makes experimental data easy to access and download, modeled after the HuBMAP Data Portal⁶. The cross-consortium Human Reference Atlas (HRA) Portal (https://humanatlas.io) makes reference atlas data and code freely available in support of data harmonization, mapping, and multiscale data exploration. All data is shared according to FAIR principles (Findable, Accessible, Interoperable, and Reusable) to maximize data utility.⁷ The HRA and SenNet data portals are synergistic: a reference atlas cannot be constructed without access to high quality experimental data and the atlas is used to harmonize and federate experimental data (e.g., spatially register tissue in a 3D common coordinate framework; to design and use standardized and validated antibody panels; to update ontologies to include novel cell types) so it can be compared across teams, organs, or assay types.

SenNet united cellular senescence researchers within the program and beyond to agree on key terminology, see Box 1. Additional terms are defined in the Getting Started with Data Submission guide⁸ and the HRA Glossary.⁹

Box 1. Key SenNet Terminology.

• Common coordinate framework (CCF): Consists of ontologies and reference object libraries, computer software (e.g., user interfaces) and training materials that (1) enable biomedical experts to semantically annotate tissue samples and to precisely describe their locations in the human/mouse body (“registration”), (2) align multi-modal tissue data extracted from different individuals to a reference coordinate system (“mapping”), and to (3) provide tools for searching and browsing experimental data at multiple levels from the whole body down to single cells (“exploration”).¹⁰

• Data curators: Curators work with a funded component’s designated data submitter to ingest data according to SenNet standards.

• Dataset: The files and metadata that comprise and describe a run of an experimental assay. Primary datasets contain the raw data and primary processing while derived datasets contain additional assays output from processing primary datasets (e.g., for spatial assays, there may be histology and -omics derived datasets)..

• Data submitters: Submitters work closely with curators to upload experimental data via the SenNet Ingest Portal.

• Data upload: One or more datasets generated by the same primary assay type submitted together to the Consortium Organization and Data Coordinating Center (CODCC).

• Metadata: ‘Data about data’ refers to descriptions about sources, tissue samples, and datasets. Metadata for a particular purpose can be defined by a metadata schema. SenNet and HuBMAP co-develop and use the same schemas (except for an additional murine metadata schema for SenNet), and modifications can be made by either consortia’s Data Coordination Working Groups.

• Provenance: The origins of each entity, captured internally via a graph database. A typical flow of provenance in SenNet is source -> organ -> tissue block -> a further block, section, or suspension -> dataset.¹¹ Everything from organ down is registered to a direct ancestor (organ registered to source (individual human or murine organism), tissue block to organ, section/suspension/block to first block, dataset to sample). A dataset is descended from the most granular sample against which the experimental modality was run.

• Reference atlas: A comprehensive, high-resolution, three-dimensional, multiscale atlas of all the cells in the healthy human/mouse body. The Human Reference Atlas (HRA) provides standard terminologies and data structures for describing specimens, biological structures, and spatial positions linked to existing ontologies.¹²

• Salutogenesis: The active production of health, derived from Latin roots meaning “the origin of health”.¹³

• Secondary senescence: The induction of senescent states in neighboring cells through SASP-driven signaling or environmental alteration.

• Senescence-Associated Secretory Phenotype (SASP): Phenotypic expressions of senescent cells that reshape tissue microenvironments and influence nearby cells.

• Senescence microenvironment (SME): Localized zones of heightened inflammation, tissue dysfunction, and declining cellular self-repair.

• Senescent cells: Cells that stop dividing but do not die; accumulate with age and stress and release harmful inflammatory substances (SASP) that damage nearby tissues.

• Senolytics: A class of drugs or compounds designed to selectively clear out senescent cells.

• Senotype: Refers to a “context dependent” senescence phenotype.

• Subclinical senescence: The early and potentially invisible stages of cellular aging that involve histologically normal cells exhibiting subtle senescence signatures that may be detectable through high-resolution assays.

• Tissue-specific aging pathways: The distinct molecular and cellular processes that drive age-related decline in different tissue types.

• Validation: The process by which curators check a data upload’s metadata (according to a metadata schema), directory and file structure (according to a directory schema) for compliance with CODCC standards.

Results

As of January 2026, SenNet results comprise (1) a set of well-defined user needs; (2) a scalable hybrid cloud microservices architecture; (3) data obtained through well-defined data submission and ingestion procedures; (4) generalized data processing and analysis pipelines; (5) reference atlas construction and usage workflows; (6) senescence focused data visualizations and user interfaces; and (7) documentation and demonstration resources. All seven contributions are detailed subsequently together with five SenNet usage scenarios that demonstrate how SenNet data and tools can be used to advance basic research and clinical practice.

General User Needs

To better understand how the SenNet Data Portal supports real-world research, we conducted ten one-on-one interviews with users and data contributors. Participants included researchers, clinicians, and computational biologists. These semi-structured conversations offered insights into user goals, workflows, and challenges. They highlighted both the strengths of the current SenNet Data Portal and opportunities for enhancement. Drawing from these interviews, we identified seven representative user stories that reflect key use cases across three domains of discovery, reference, and translation, see Table 1. Each story outlines how the Portal adds value—whether through data exploration, validation, tool development, or cross-species comparison—and helps illustrate the diverse ways users engage with SenNet resources in their work. Additionally, we conducted follow-up interviews with four researchers in order to understand their Portal workflows in more depth. These context-rich, illustrated Usage Scenarios are presented in the Results section.

Flexible hybrid cloud microservices architecture

The SenNet Consortium Organization and Data Coordination Center (CODCC) developed a flexible hybrid cloud microservices architecture^3,4 based on the HuBMAP Consortium^3,4 and data portal to support data curation, ingestion, integration, access, analysis, exploration, and download (https://portal.hubmapconsortium.org). See cloud microservices architecture details in a related HRA publication¹⁴ and flow of data through Portal architecture from data submission to consumption in the HuBMAP Portal paper. This microservices architecture was adopted for the SenNet Portal (https://data.sennetconsortium.org) on distinct infrastructure, demonstrating the approach is transferable among supercomputer centers, public cloud, university infrastructure, and mixed/hybrid deployments. Data is ingested using the Ingest UI (Fig. 1, e.g., the Ingest UI is made up of parts of the Portal Application, Ingest API, and Ingest Pipeline, while leveraging other APIs). The HRA Registration User Interface (RUI)^10,15,16 supports tissue registration in 3D so tissue blocks can be explored within the 3D human reference body using the Exploration User Interface (EUI)¹⁰. Data is then processed and analyzed using diverse cell segmentation and cell type annotation tools, including Azimuth¹⁷ and made available via the Portal. Vitessce¹⁸ supports the exploration of multi-modal spatial and single cell data⁶, see example in usage scenario section Hypothesis Generation.

Figure 1: SenNet infrastructure architecture and functionality.

On premises resources at the University of Pittsburgh and public cloud resources in AWS are used to store, ingest, process and expose the data publicly via the SenNet Data Portal. APIs exist for all programmatic discovery and access to datasets. The API Gateway monitors all API access and allows only authorized access to SenNet provenance and metadata. Applications support data ingest, analysis, visualization, search, and download via diverse user interfaces. Globus Auth is used for all authentication and authorization to non-public SenNet APIs, applications, and data. All data is available for download via Globus Collections. Data upload and download are managed via Globus Transfer.

Major advances of SenNet beyond the HuBMAP data portal include: The SenNet Data Portal integrates ingest functionality into the public data portal (HuBMAP has separate Ingest and Data portals). RUI-based tissue registration is required for all human SenNet data (in HuBMAP, 3D block registration is encouraged). SenNet supports mouse and human sources, including senescence-induced and diseased (HuBMAP contains healthy human tissue). SenNet supports additional tissues (e.g., muscle, adipose, mammary gland) and improved navigation and visualization of provenance information. Provenance and data flow visualizations are directly interactive for investigative feedback, orientation to data and provenance, and search in SenNet, using the framework of data-driven documents, substantially enhancing user experience.

Data Submission and Ingestion

Data submission and ingestion in SenNet is curated by the CODCC. Submission begins with a bulk data upload by a data submitter. Ingestion is the process of the curator handling data submitters’ bulk data uploads via validation, reorganization, central analysis, quality assurance, and publication of the data and metadata. SenNet adopted HuBMAP’s workflows, metadata and directory schema¹⁹, and co-developed new schemas. SenNet has also developed a unique mouse source metadata schema²⁰ since it supports both human and mouse tissue samples.

Datasets submitted to the CODCC within bulk uploads require at least two metadata files, one describing the assay and the other recording applicable contributor information. Data is organized according to the directory and file structure described in the relevant directory schema for the assay. Preparing an upload through data publication involves collaboration among curators and data submitters. Avenues for collaboration include weekly SenNet Data Coordination Working Group (DCWG) meetings to set standards and share pragmatic approaches throughout the process, twice-a-week curator office hours, and triage in cooperation with the SenNet Help Desk. Curators and submitters work together to publish experimental protocols on SenNet’s protocols.io workspace²¹ (218 distinct protocols to date), register entities in the provenance chain (sources, organs, and tissue samples), prepare and ingest metadata, and submit spatial coordinate information for human tissue blocks registered in the HRA’s Registration User Interface.¹⁰

Data Processing and Analysis

The CODCC includes focus on data ingestion, analysis, and visualization workflow integration and tool development, development of Azimuth references²², and the Human Reference Atlas²³ in close collaboration with external experts. The SenNet portal hosts datasets processed through uniform computational pipelines, one per assay type. This includes a single-cell/single-nucleus transcriptomic quantification pipeline built on the Salmon method²⁴ and ScanPy for downstream analysis²⁵, single-cell quantification of open chromatin regions (scATAC-seq) built on ArchR²⁶, and cell and nucleus segmentation of multichannel images performed by the top-performing segmentation method for each particular assay²⁷, then unsupervised downstream analysis via the SPRM package²⁸. When possible, uniform processing pipelines include senescence-specific analysis, such as the inclusion of DeepScence²⁹ into the sc/snRNA-seq pipeline, producing predictions of cellular senescence for all sc/snRNA-seq datasets that have been processed since the inclusion of this pipeline step. This uniform processing of datasets between providers, assay variants, and tissues allows for easier integrative analysis, removing artificial barriers to data harmonization, and supports the senescence analysis detailed in other sections of this paper.

SenNet Atlas Construction and Usage

Spatially registered and ontology-aligned SenNet data is used to construct a reference atlas in support of senescence studies, see Box 1 for terminology. In January 2026, there exist 1,927 spatially registered SenNet datasets for 864 tissue sections, cut from 566 tissue blocks with 425 extraction sites in 11 organs from 219 donors, uploaded by 7 tissue data providers, see Exploration User Interface (EUI, https://apps.humanatlas.io/eui)10 with SenNet filter applied. This all-HRA EUI does include secondary datasets, (e.g., cell by gene matrices or cell type annotation and segmentation masks) computed from the primary data; and it uses data from the HRA KG which is updated every six months with the HRA release. The all-HRA EUI provides access to a total of 7,281 tissue datasets, with 1,770 (or 24.3 percent) coming from SenNet.

SenNet’s EUI at https://data.sennetconsortium.org/ccf-eui only includes primary datasets, analogous to counts on the SenNet Data Portal. It supports easy access to 1,042 public datasets from 194 donors. With consortium login, 1,375 datasets from 361 donors are accessible and will become available publicly after data validation has been completed.

Exactly 54 of these datasets were used to compute HRA cell type populations (HRApop) v1.0 for anatomical structures³⁰ in June 2025 (52 from the lung and two from the liver). The other 988 datasets were either not in existence when HRApop v1.0 was computed or came from organs not covered by existing cell type annotation tools (e.g., 86 datasets from placenta and 65 datasets from the ovaries). That is, 8.1% of the HRApop v1.0 data comes from SenNet, making it a major data source for HRApop construction. Using HRApop, cell type populations and biomarker expression values can be predicted for HRA-registered 3D tissue volumes—before any single cell assay type is run; inversely, given a cell type by biomarker expression matrix, the 3D volume from which this tissue was cut can be predicted.

Five CODEX human lymph node datasets from SenNet with 8,918,854 cells were used in a recent cross-consortium study of spatially resolved omics data that explores endothelial cell environments across 14 studies comprising 12 tissue types and a total of 47,349,496 cells³¹ All data and code is freely available and the Cell Distance Explorer (CDE, https://apps.humanatlas.io/cde) has been used to analyze other SenNet human and mouse datasets.

Data Visualizations and User Interfaces

Different user interfaces exist to explore experimental tissue datasets and reference atlas data.

User interfaces to experimental data

Vitessce:

Visualization Tool for Exploration of Spatial Single-Cell Experimentation (Vitessce) is an open-source, web-based framework designed for the integrative visualization of multi-modal, spatially resolved single-cell data in both 2D and 3D (https://vitessce.io). It enables researchers to interactively explore high-resolution imaging supported by Viv³², spatial transcriptomics, and conventional single-cell data in a linked visualization interface. A diverse set of Vitessce visualizations, spanning spatial and non-spatial as well as single- and multi-modal data, was originally developed for the HuBMAP Data Portal to support specific assay requirements and analytes across 16 data types. These visualizations are also fully supported by the SenNet Data Portal, see the Hypothesis Generation section for exemplary Vitessce visualizations used with SenNet data.

User interfaces to reference atlas data

There are diverse user interfaces (UIs) that support atlas construction (e.g., authoring ASCT+B tables or OMAPs; spatial and semantic registration of tissue blocks) and UIs that support atlas usage (e.g., predicting cell type populations for registered tissue blocks). We explain the former first, realizing that few experts will construct the HRA, most experts will use the HRA.

ASCT+B Reporter with Organ Mapping Antibody Panel (OMAP) Support:

The ASCT+B Reporter supports the visualization of ASCT+B tables³³ in the form of trees showing anatomical structures (AS) in the human body, the cell types (CT) located in them, and the biomarkers (B) used to characterize these cell types (apps.humanatlas.io/asctb-reporter). A mode was added to explore, compare, and optimize OMAPs³⁴ specific to organs, multiplexed antibody-based imaging methods, and tissue preservation methods. The ASCT+B Reporter is critical for authoring, reviewing, and optimizing ASCT+B tables and OMAPs.

SenNet-branded Registration User Interface (RUI)¹⁰:

The SenNet Data Portal (data.sennetconsortium.org) enables data providers to register data sources (donors), organs, tissue blocks, etc. A RUI instance, embedded as a web component, is available during this registration workflow to indicate the spatial origin (also called extraction site) of a tissue block inside one of the 73 3D reference objects in HRA v2.3 representing the organs of the healthy, adult human body. Details are available in the data submission guide.³⁵

SenNet-branded Exploration User Interface (EUI¹⁰):

Enables the spatial and semantic exploration of SenNet donors, tissue blocks, tissue sections, and datasets if they have been annotated with an extraction site via the RUI. Each extraction site has a 3D position, rotation, and size. The SenNet specific EUI is available at data.sennetconsortium.org/ccf-eui, the all-HRA EUI is at apps.humanatlas.io/eui. See example in usage scenario section Cell Distance Distributions.

Cell Distance Explorer (CDE³¹):

Enables a user to visualize cell-to-nearest-anchor-cell linkages (apps.humanatlas.io/cde) and distance distributions. The user provides a CSV file with a list of cells (x, y, z-position, plus a cell type label) and metadata (e.g., visualization title, imaging technology, organ, sex) and designates an anchor cell type (e.g., endothelial cells). The CDE then computes links between each cell type and its nearest anchor cell. As auxiliary visualizations, the CDE provides a histogram of distances, color-coded by cell type (and a bucket for all), as well as one violin plot per cell type with an inline box-and-whisker plot. The CDE allows researchers to analyze and compare changes in cell-cell distance distributions across tissue regions, between normal and diseased tissue, and across diseases. It thus enables exploration of how cell organization is impacted by different conditions. See example in usage scenario section Cell Distance Distributions.

Cell Population Graphs (apps.humanatlas.io/cell-population-graphs):

Make it possible to compare cell type populations across experimental datasets and in relation to HRApop atlas datasets. Each stacked bar graph presents an experimental or atlas dataset; different cell types are rendered in different colors and bar height corresponds to cell count or percentage. Bars can be sorted by total cell count or prevalence of specific cell types, and they can be grouped by dataset-specific properties (e.g., sex, developmental stage, ethnicity if metadata is available). A newer interface is the HRApop Visualizer (apps.humanatlas.io/hra-pop-visualizer), which enables exploration of cell type populations for anatomical structures, extraction sites, and datasets in HRApop v1.0.³⁰

HRA Organ Gallery virtual reality (VR):

A stand-alone VR application (http://humanatlas.io/hra-organ-gallery), available free of charge in the Meta Store.³⁶ Users can view 77 3D reference objects and 2,638 tissue blocks linked to 8,045 datasets from 724 donors across 18 consortia—including 566 tissue blocks, 219 donors for SenNet but also data from the Human BioMolecular Atlas Program^3,4 (HuBMAP) and the Genotype-Tissue Expression (GTEx).³⁷ Additionally, two SenNet datasets (CODEX, Visium) can be explored in immersive 3D visualizations.

Documentation and Demonstrations

The SenNet portal (https://sennetconsortium.org) provides five main dimensions of engagement: resources, data exploration, news, involvement, and member services.

The Resources section provides the technical backbone for contributing and using SenNet data. Within this section, users can access biomarker lists, documentation of software tools, the Human Reference Atlas, experimental protocols on protocols.io, SenNet publications in PubMed and via Google Scholar profile with citations, software on GitHub (https://github.com/sennetconsortium), as well as summaries of novel technologies and tools developed within SenNet.

Data forms the core of the portal, organized around both the structured submission of and interactive access to datasets. It first supports contributor workflows, offering clearly defined entry points for registering specimens, uploading data, and validating metadata through standardized templates. These processes are guided by detailed documentation, and contributors can monitor submission status to ensure compliance with consortium-wide quality controls. The same section also provides access to curated and published datasets. New users can begin with onboarding materials, explore interactive data previews through the dynamic SenNet Sankey diagram (https://data.sennetconsortium.org), or browse metadata organized by popular search categories. More experienced users can search and filter datasets by tissue, assay, contributor, or organ, view integrated summaries and visualizations, and export results for downstream analysis. In addition to web-based interfaces, the portal supports programmatic access through Application Programming Interfaces (APIs). The Usage Scenarios section illustrates how these tools can be combined for data exploration, integration, and reuse across diverse research contexts.

News chronicles the consortium’s ongoing activities, offering a view of both internal milestones and public-facing achievements. Recent news items have included announcements (calls for reviewers, fellowships, training sessions), events (conferences, workshops, and symposia), funding opportunities, and new publications³⁸. The portal’s interview series features conversations with scientists and project leaders that trace how different efforts within SenNet connect and progress over time (https://sennetconsortium.org/interview-series). The SenNet Sentinel is a monthly newsletter spotlighting announcements, publications, programs, and personnel for the consortium’s internal and external audiences.

The Involvement section defines two major pathways through which researchers and institutions can take part in SenNet’s collective work. They can either apply to become Associate Members of the consortium (there are eight in January 2026), contributing to its research goals directly. Alternatively, users can join working groups dedicated to benchmarking, data coordination, atlas construction, or emerging technologies, allowing them to participate in shaping the methods and standards that guide SenNet’s evolving research practices.

Member services form the administrative core of the site facilitating internal communication and coordination between consortium members. Through a secure login, consortium members can access their user profiles, upload and review experimental data so it can be published, follow internal calendars, share meeting materials, and documents.

Usage Scenarios

General user needs were detailed in the General User Needs section. Here, we present five exemplary data and tool usage scenarios that showcase how SenNet data can be used to (1) benchmark and improve noise cleaning algorithms and tools, (2) showcase that pilot transcriptomics lung data in mouse is supported by standardized, consortium-generated datasets, (3) guide pathologists in spatial and other data analyses in diabetes in human pancreas with a focus Islets of Langerhans functional tissue units (FTUs), (4) run a temporal metastudy, and (5) compute cell distance distributions for immune cells.

Data Quality Benchmarking

A researcher focused on transcriptomic instability and cellular senescence is grappling with a long-standing challenge in the field: senescence is not defined by a single biomarker, but by multiple biomarkers and signatures that are highly cell type and context specific (see subclinical senescence in Box 1). Differentiating these weak signals from other biological signals and technical noise requires exceptionally clean data. Traditionally, transcriptomic datasets compiled across multiple studies have suffered from strong batch effects arising from differences in protocols, lab environments, or sequencing technologies.³⁹ Researchers have often had to resort to crude normalization methods that risk erasing precisely the subtle differences in biological signals they hope to study, including variations in senotypes. SenNet’s carefully curated data, generated in a coordinated manner across a small set of labs using standardized protocols, offers something rare: large-scale, high-quality transcriptomic data with low technical noise. For a meta-study involving multiple organs, Bartz and colleagues tapped into SenNet’s harmonized raw datasets, using them both to benchmark existing noise-measurement algorithms and to train data-hungry deep learning models capable of distinguishing emerging senescence micro-environments. They benchmark methods such as Global Coordination Level Analysis (GCL)^40,41 and Scallop⁴² for transcriptional noise, alongside senescence detection tools like hUSI⁴³, SENCAN⁴⁴, and SenCID.⁴⁵

Hypothesis Generation

Pulmonary biomedical researchers preparing resubmission of a major grant turns to the SenNet Portal to strengthen their hypothesis about senescence markers in lung tissue. The Portal’s value lies not only in providing access to high-quality, standardized, multi-modal datasets but also in making them quickly explorable through a user-friendly interface at https://data.sennetconsortium.org/search. After filtering for published datasets, Drs. Lee and Kim select mouse lung transcriptomics data and use the built-in Vitessce UMAP visualization and gene expression panels to test candidate genes like Cd74, identifying clusters of B cells with high Cd74 expression (Fig. 3a).

Figure 3. Visual analysis of gene expression in mouse lung tissue.

a. The SenNet Portal supports interactive Vitessce visualizations of human and mouse datasets. Shown here is RNA-seq data from the left lung of a 5-month-old female C57BL/6J mouse. After searching for the candidate gene Cd74, the user can visualize its expression across all 9,700 cells in a UMAP plot (top left), heatmap (bottom), and by cell type in a violin plot (right). The Expression by Cell Set panel shows high Cd74 expression in B cells. (SenNet ID: SNT243.LTGW.532). b. The Vitessce lasso tool enables interaction with the UMAP plot and supports creating user-defined cell sets. The user can select cell regions that express Cd74-high (blue) and Cd74-low (teal). These new groups appear in the Cell Sets panel, where they can be managed (e.g., renamed). The violin plot visualizes the differential expression of Cd74 between the user-defined sets. (SenNet ID: SNT243.LTGW.532). c. The user can search for specific genes (e.g., Hmgb1) in the Gene List panel on left and use the Expression by Cell Set violin plot to compare Hmgb1 expression values for custom cell sets (e.g., SenNet ID: SNT243.LTGW.532).

The Vitessce lasso tool (Fig. 3b) allows researchers to define custom groups–e.g., Cd74-high and Cd74-low–and to compare them for expression of senescence markers such as CDKN2A or MHC class II genes. The preliminary evidence they gain from this comparison reinforces the original hypothesis, which they extend by comparing young (5-month) and old (23-month) mouse lungs and contrasting male and female samples, using datasets from the same upload batch to probe whether the candidate markers participate in tissue-specific aging pathways or reflect broader aging signatures across cell types. They can also compare data across modalities—checking protein-level expression in antibody-based imaging datasets—to see whether transcript-level patterns are reflected at the protein level. These comparisons and results give the researcher the confidence to argue that findings in preliminary data are supported by standardized, consortium-generated datasets. Next, a user can search for other senescence-related genes (e.g., Hmgb1) in the Gene List panel and use the Expression by Cell Set violin plot to compare Hmgb1 expression between user-defined groups (Fig. 3c). Specifically, the observation that Hmgb1 shows enriched expression within the Cd74-high subpopulation suggests a potential functional synergy between these two molecules in the mouse lung. Given that both HMGB1 and CD74 are interrelated mediators of inflammation, this data-driven insight empowers researchers to formulate new hypotheses, such as investigating whether concurrent blockade of these targets offers synergistic therapeutic benefits in lung injury models.

Spatial Analysis

In an aging and senescence research lab, biomedical and computational scientists collaborate to understand how human pancreatic function declines with age and disease. Using the SenNet Portal, the team retrieves high-resolution, multi-modal pancreas datasets—H&E histology, spatial transcriptomics (Xenium), and protein imaging (CODEX)—from healthy donors. The pathologists guide biological questions about insulin production, while computational researchers develop Jupyter notebooks that overlay these data layers, segment cells, and apply spatial clustering algorithms to detect islets of Langerhans as functional tissue units (FTUs).⁴⁶ Together, they quantify islet composition, geometry, and proximity to ducts or vasculature, revealing which islets maintain healthy beta-cell function and which show signs of senescence or dysfunction. This integrated workflow enables the lab to link spatial abnormalities to early markers of pancreatic cancer. For example, does secondary senescence propagate across neighboring islet cells? It also provides a generalizable framework that can apply to FTU detection in other organs through SenNet’s multi-layer datasets. Data and code developed as part of the Human Reference Atlas effort, including 2D illustrations of FTUs⁴⁶, FTU segmentation algorithms^46,47, and Cell Distance Explorer analyses³¹, support this effort and make it possible to explore SenNet data and results in the HRA.

Cost-efficient Profiling

A researcher interested in human lung aging without access to large experimental budgets or extensive human tissue banks uses the SenNet Portal to study how cell type composition shifts across the lifespan. By querying the Portal’s APIs, they retrieve all published single-nucleus RNA-sequencing datasets derived from human lung (Fig. 4a) and pass them through a custom processing pipeline, integrating the data using open-source Python modules, including Vitessce visualizations.^25,48–54 They then use annotation tools, like Azimuth¹⁷, CellTypist^55,56, and other cell type annotation tools available via the HRA to harmonize cell identities across datasets (Fig. 4b). Grouping donors into age brackets reveals how the proportions of alveolar epithelial cells, endothelial cells, and immune populations change across the lifespan, offering hints of potentially subclinical aging pathways at work before overt conditions emerge (Fig. 4c). The researcher can further extend this analysis to control for other donor factors—such as BMI, blood type, medical/social history—and compare changes in cell type composition across other organs to determine whether age-related shifts are organ-specific or systemic. The result is an accessible and cost-effective way to track aging at both the cellular and cross-organ level without duplicating costly profiling efforts. Using atlas-level consortium data to better understand temporal lung cell population dynamics, the researcher can design more focused pilot studies to explore the underlying causes and effects of such changes.

Figure 4. Integration reveals changes in human lung cell type composition across lifespan.

a. UMAP embeddings for each of the 66 single-nucleus RNA-seq datasets may be visualized with software such as Vitessce withinSenNet data portal functionality. b. Custom code is used to retrieve, filter, and integrate data across donors. Resulting harmonized CellTypist annotations are visualized on a single UMAP plot. (c) Donor metadata is used to group datasets by age quartile, demonstrating how proportions of putative epithelial, endothelial, and immune cells are influenced by donor factors such as age.

Cell Distance Distributions

Researchers interested in understanding immunology can use the HRA Cell Distance Explorer (CDE)³¹ to understand the location and distances between different cell types. Fig. 5 shows one lymph node dataset with 1,826,800 segmented and annotated cells from Fan’s lab for a 25-year old female donor (SenNet Dataset ID: SNT584.CWSK.568). Data was obtained using Formalin-Fixed Paraffin-Embedded tissue samples cut from a lymph node in the stomach (Fig. 5a). A 49-plex marker panel optimized to identify B cells, T cells, macrophages, and other immune cell types was used to image the sample using PhenoCycler Fusion 1.0. The CDE was used to compute distances from all cell types to the nearest endothelial cells, which are then visualized as a violin plot showing distance distributions for each cell type (Fig. 5b). The CDE shows an interactive view of the data with cell-to-nearest-endothelial cell linkages allowing further inspection and closer examination of the data, see linkages in Fig. 5c that are colored by cell type. Examining the violin plot in the CDE shows the difference in absolute spatial location of different cell types to the endothelial cells, especially the different immune cells, i.e., the subtypes of B cells and T cells. The visualization also shows whether the spread of a cell type around the blood vessels is tighter or varied by looking at the interquartile ranges in the violin plots. The researcher can further examine changes in the distance distributions for samples from different donors, looking at changes in spatial cell organization with respect to certain anchor cells across age groups, sex, conditions, and regions.

Figure 5. Cell distance distributions for immune cells in a lymph node of a 25 year old female donor.

a. Lymph node tissue RUI-registered and shown in the context of the HRA female reference body. b. Violin plot of cell distance distributions for different color-coded cell types. c. Cell-to-nearest-endothelial cell distance visualization in the CDE interactive viewer.

Across donors spanning age, these distance distributions can reveal age-associated remodeling of vascular-associated lymph nodes. Spatial architecture of lymph node tissues in older donors are often described by a smaller node size, capsule thickening, fibrosis, lipid and hyaline deposition, and loss of clearly segregated B cell and T cell zones⁵⁷. Furthermore, increased lymphatic vessel permeability and altered morphology of HEVs may further reshape perivascular localization patterns which results in altered immune cell entry. In the context of immune aging, such trends can be interpreted as spatial signatures of immunosenescence⁵⁸. The changes in endothelial niches reshape where immune cells reside and impact antigen surveillance of functional microenvironments such as follicular and T-zone niches.

In malignancies of lymph nodes, such as Angioimmunoblastic T cell lymphoma (AITL), there is significant remodeling of the normal lymph node architecture, characterized by infiltration of diverse immune cells, extensive follicular dendritic cell meshwork proliferation and prominent arborizing high endothelial venules (HEVs)⁵⁹. Specifically, T follicular helper cells (Tfh) which are the tumor cell of origin are found in close proximity to the HEVs, highlighting deviations in cell distance distributions relative to healthy lymph nodes⁶⁰.

There are a total of seven lymph node datasets that have been registered in the female reference body so they can be explored in the context of other organs in the SenNet EUI, go to https://data.sennetconsortium.org/ccf-eui and filter by female and Lymph Node. The CDE Python template was used to compute the cell-to-nearest-endothelial cell distance distributions.

Usage Statistics

The SenNet-internal Data⁶¹ Ingest Board (Fig. 6a) lists the status of all 5,061 datasets (4,295 primary plus derivative datasets) that were uploaded whether or not they have been processed or published.

Figure 6. SenNet Dashboards.

a. The internal Data Ingest Board lets SenNet members sort and filter datasets by authoring team, status, type, and 20+ other properties. b. The HRA Usage Dashboard shows HRA digital object requests from the SenNet Portal since June 2023, computed daily with a two week rolling average.

Usage for the SenNet Consortium public Web Portal (https://sennetconsortium.org), Data Portal (https://data.sennetconsortium.org), related sites and services (https://docs.sennetconsortium.org), is logged via AWS, Google, and GitHub. During the two year period of February 1, 2024 to February 1, 2026, from Google Analytics, the Web Portal has been viewed by 21,700 active users, the Data Portal viewed by 7,500 active users, and for the Documentation and Biomarkers site viewed by 3,200 active users, each increasing over time. Top referrers include: Google, Bing, trafficcheap.com, leadsgo.io, the-scientist.com, and nature.com. Most viewed pages include the base page (/), /search, /dataset, and /biomarkers with more than 5000 Data Portal users from the United States, and next more than 1,000 from China.

Also in the last two years, custom AWS logging of Downloads, APIs, and services reveals over 68.45 TiB of data has been downloaded, and more than 6,079,000 API calls have been made (including both direct use of published APIs via SDK/SmartAPI and those same calls that operate the portals). Top services include /search, /datasets. Top referrers are direct, auth.globus.org, and google.

GitHub logged consortium contributions and usage for the infrastructure alone (not including individual site tools and algorithms) include over 40 repositories that have been viewed 4,600 times and cloned 5,300 times. See: https://github.com/x-atlas-consortia/

For the HRA Portal, between June 2023 and end of January 2026, 3,316,899 HRA Portal requests and 1,693,128 HRA API requests were fulfilled; the top-5 referrers are the GTEx Portal (https://gtexportal.org), the HuBMAP Entity API, the HuBMAP Data Portal, the SenNet Data Portal (https://data.sennetconsortium.org/search) (Fig. 6b), and EMBL-EBI (https://www.ebi.ac.uk). The 3D reference objects were accessed 34,704 times via the NIH3D website. The HRA OWL file was accessed 28,266 times via the NCBI BioPortal Ontology Browser (https://bioportal.bioontology.org/ontologies/HRA) and 39,577 times via the EBI OLS Ontology Browser (https://www.ebi.ac.uk/ols/ontologies/hra). 310 students registered for the VHMOOC and spent 5,652 hours reviewing materials, taking self-tests, and engaging in a community of practice.

In total these reveal robust usage of information, data and services of the SenNet Consortium.

Discussion

This Resource paper describes data, code, and tools that are of broad utility, interest, and significance to the senescence research community. Five usage scenarios exemplify how SenNet data and tools can be used to benchmark algorithms, generate hypotheses, perform spatial analyses, run cost efficient profiling, and explore senescent biomarkers in 3D. We conclude with a discussion of limitations and planned extensions.

In January 2026, 1,753 primary datasets of the 5,061 total primary datasets were publicly available in the SenNet portal. The remaining datasets are either incomplete (e.g., in progress, uploading metadata, etc.) or are undergoing QA/QC protocols prior to publication. Experimental teams closely collaborate with the CODCC curators to finalize and publish these datasets via the SenNet Data Portal so they can be used by the worldwide scientific community.

We plan to make the usage scenarios presented here and others developed by SenNet members easy to access via the SenNet Data Portal and the HRA Portal. The usage scenarios help highlight unique data assets, provide hands-on guidance on how to use existing APIs and user interfaces, and showcase custom code developed by experimental teams.

The SenNet Portal will soon feature more “publication pages”. Publication pages highlight peer-reviewed publications and preprints that use consortium datasets.⁶² They make it easy to access and download at once, all datasets and code used in a paper. Two SenNet publication packages are planned in the Cell Press and Nature portfolio of journals. Papers in these packages are expected to have publication pages with a summary of publication-related information, a list of referenced SenNet datasets, plus relevant visualizations.

We use the SenNet Data Portal Usage data and dashboards for understanding and optimizing data and tool usage. Current dashboards (Fig. 6a) are only available to SenNet members, but a public dashboard showing the number of users and accesses over time is planned.

In parallel to the work presented here, the SenNet CODCC team is working on expanded spatial-omics releases and improved tools for senotype characterization by the consortium. This will enable assertional knowledge developed by the consortium to integrate fully in the same graph database as dataset metadata for further discovery and validation of senescence science.

Methods

Data Quality Benchmarking (Josh Bartz)

A single-cell RNA-seq (scRNA-seq) computational pipeline was developed for processing and benchmarking the datasets. The pipeline processed raw scRNA-seq datasets through a 7-step workflow: Cleaning, Integration, Clustering, Cell Type Identification, Calculating Scallop scores, Calculating Global Coordination Level (GCL), and Visualization. The workflow allowed comparison of different methods of measuring transcriptional noise. The following datasets were used by this pipeline to generate Fig. 2: SNT246.PGHH.774, SNT449.GRWG.598, and SNT397.SPGW.436. Seurat was used for data cleaning, integration, clustering, and cell type identification (steps 1–4). Since GCL calculations can be computationally demanding, the overall GCL analysis was split into smaller chunks which was then run on high performance computers (HPCs) using SLURM job scheduler. The complete code (predominantly in the R programming language) for all seven steps is available via GitHub (see Code Availability).

Figure 2. Benchmark comparisons of Global Coordination Level and Scallop.

The high-quality lung transcriptomic dataset provided by SenNet facilitates robust benchmarking of algorithms, including those designed to quantify transcriptional noise, across a broad range of cell types. Processed data from following SenNet datasets were used in this example: SNT246.PGHH.774, SNT449.GRWG.598, and SNT397.SPGW.436.

Spatial Analysis (Sergii Domanskyi)

A set of 12 Jupyter notebooks demonstrating data access, analysis, and visualization workflows for the SenNet Data Portal were created (see Code Availability). These notebooks showcase practical applications of spatial analysis across computational biology, with a focus on tissue architecture, cell-cell interactions, and disease-related cellular changes for high-resolution, multi-modal datasets including spatial transcriptomics, imaging, and single-cell RNA sequencing data from diverse human and mouse tissues. The demos also provide accessible entry points into senescence research for discovering datasets, performing gene expression analysis, and integrating multi-modal imaging data. The raw and processed datasets used throughout these notebooks were retrieved from the SenNet Portal. These include datasets SNT576.PXPP.452, SNT484.VLRN.777, SNT443.KFNS.239, SNT574.MMQC.683, SNT594.JCFN.856, and SNT227.HLMG.672 from collection SNT793.SZRS.468, dataset SNT544.XHGB.538 from collection SNT947.NXPB.793, and datasets SNT268.LXPG.784, SNT324.BDTT.263, and SNT469.JJLS.674 from collection SNT566.NMTV.379.

Cost-efficient Profiling (Sam Peters)

Relevant single-cell datasets were queried using the parameterized search endpoint of the SenNet Search API, and expression matrix files of qualifying datasets were downloaded using the Entity and Assets APIs. Datasets were then processed and integrated with Python (v3.9.23) using scanpy (v1.10.3)⁴² and anndata (v0.10.9)⁴⁰. Basic quality control metrics across cells and genes were calculated. Then, basic thresholds were set to filter out noisy data: cells that had fewer than 500 transcript counts; cells with more than 10% mitochondrial content; and genes expressed in fewer than 3 cells across samples. Data were normalized and log1p transformed prior to scaling for dimensionality reduction. Highly variable genes were annotated in a batch-wise manner, followed by scaling and PCA. Data was then integrated to adjust the principal components (PC) using Harmony (v0.0.10)⁴⁶, and 30 integrated PCs were used to calculate 30 nearest neighbors. Finally, a UMAP⁶³ embedding was computed for visualization and CellTypist (v1.7.1)^47,48 was applied to annotate cell types based on the Human Lung Cell Atlas (HLCA, v1.0) model. The processed data was used to examine the effect of age on cell type composition. All code is publicly available (see Code Availability). Processed data from following SenNet datasets were used in this usage scenario example: SNT246.PGHH.774, SNT397.SPGW.436, SNT449.GRWG.598, SNT498.XKRG.893, SNT233.XMTD.574, SNT722.MQMW.937, SNT673.HKBD.656, SNT759.KMJC.586, SNT627.TGCJ.324, SNT985.MXCW.829, SNT266.DBBP.698, SNT493.RSKL.732, SNT776.QQGW.623, SNT536.TGBM.294, SNT967.KBJF.497, SNT867.PMJD.853, SNT478.QGSL.332, SNT394.ZPJN.579, SNT826.HWLM.338, SNT792.KVKQ.344, SNT372.NMWW.264, SNT577.KHFG.572, SNT367.FWNH.537, SNT328.LPZQ.663, SNT657.STXR.948, SNT593.SCSG.422, SNT299.DVNG.578, SNT933.RZWH.988, SNT339.DQDB.424, SNT362.BRTK.755, SNT647.WHJN.292, SNT489.CCVQ.324, SNT829.BMNR.662, SNT794.NKKN.985, SNT669.GCNV.685, SNT864.KHJR.253, SNT479.PCMQ.829, SNT253.MKNV.455, SNT236.VDFB.848, SNT654.XSRN.726, SNT467.NJWW.933, SNT676.SXTP.348, SNT984.MSST.526, SNT773.DRXG.427, SNT945.WCZH.229, SNT497.HRKZ.873, SNT644.VGFZ.956, SNT977.VVTW.775, SNT468.XMQP.379, SNT658.LQWB.482, SNT872.FZFV.997, SNT432.BQVB.669, SNT767.MZKB.693, SNT627.QHCP.732, SNT959.SGTC.264, SNT625.CWGD.777, SNT289.MKGJ.729, SNT648.KPTH.546, SNT983.FMRG.925, SNT587.CNQX.768, SNT835.GVZP.228, SNT757.BGFN.798, SNT387.XHRC.584, SNT473.ZJKW.727, SNT887.TMFM.723, SNT493.TKJD.678.

Cell Distance Distributions

Using the Cell Distance Explorer python template at https://portal.hubmapconsortium.org/templates/hra_cde_tutorial, a Python Jupyter notebook was created that loads the cell table containing x-y coordinates and cell type assignments for a lymph node dataset of a 25-year-old female donor (SenNet Dataset ID for LN00560 sample: SNT584.CWSK.568) and visualizes the cell distance distributions for 1,826,800 cells. The cell table for this sample was processed in prior work³¹ and was used for this analysis and demo. All code is publicly available (see Code Availability) and the sample cell table used can be downloaded from the CDE Spatial Omics Gallery (https://apps.humanatlas.io/cde/gallery/lymphnode-codex-yale). For a non-code user, the same analysis can also be computed and visualized using the browser-based application available at https://apps.humanatlas.io/cde.

Table 1: User stories.

Feature summary, target user roles, user activities, and added value for seven user stories that are significant for SenNet Portal users.

Feature	User Role	User Activity	SenNet Portal Added Value
Discovery Use Cases
US#1. Exploratory Analysis	Researchers, Clinicians	Analyzing preliminary findings about gene expression and tissue architecture	View and compare gene expression and tissue structure across tissues, age groups, and species to better interpret early findings
US#2. Integrative Analysis	Researchers, Clinicians	Investigating how circulating molecular signals influence dysfunction across organs	Support systems-level research by integrating senescent cell profiles and circulating signals across tissues, enabling discovery of cross-organ mechanisms of dysfunction
Reference Use Cases
US#3. Data Validation	Researchers	Identifying senescence in particular tissues using specific markers	Use reference datasets to validate findings in unfamiliar tissues or conditions, reducing the need to generate new data
US#4. Normal Tissue Comparison	Pathologists	Examining gene expression and tissue architecture to discover early disease markers	Compare samples to histologically normal reference tissues to detect early markers of disease in at-risk populations
US#5. Tool Evaluation	Computational Researchers	Evaluating the performance of existing machine learning models to develop benchmarking tools for senescence	Use diverse, multi-tissue datasets to train and evaluate machine learning models, improving accuracy and reproducibility in senescence detection
Translational Use Cases
US#6. Therapy Development	Researchers, Clinicians	Identifying shared and tissue-specific pathways of aging	Map signaling pathways from mouse to human to identify corresponding tissues and cell types, enhancing the relevance of preclinical models
US#7. Cross-Species Translation	Researchers, Clinicians	Translating well-characterized signaling pathways from mouse to human	Identify human tissues and cell types that correspond to mouse models, improving the relevance of preclinical findings

Data Availability

All SenNet data is available via the SenNet Portal (https://portal.hubmapconsortium.org). Azimuth references can be accessed at https://azimuth.hubmapconsortium.org. HRA data and code are available at the HRA Portal (https://humanatlas.io).

Code Availability

Code is available on three different GitHub organizations: (1) https://github.com/sennetconsortium is the primary SenNet organization, (2) https://github.com/cns-iu is the organization owned by the Cyberinfrastructure for Network Science Center at Indiana University, and initial experimental HRA code starts here, (3) https://github.com/x-atlas-consortia was created to host cross-consortia code, including hra-kg, hra-pop, hra-apps, and hra-api. APIs, available as RESTful web services, support data ingest, querying, and delivery of metadata, see details at https://docs.sennetconsortium.org/apis. Code from the Usage Scenarios section is available at https://github.com/joshbartz55/SenNetSingleCellRNAseqDemo (Data Quality Benchmarking), https://github.com/TheJacksonLaboratory/sennet-demo (Spatial Analysis), https://github.com/TheJacksonLaboratory/sennet-demo/blob/main/notebooks/demo-snrna-seq-all-human-lung.ipynb (Cost-Efficient Profiling), and https://github.com/x-atlas-consortia/hra-notebooks/blob/main/notebooks/hra-cde-sennet.ipynb (Cell Distance Distributions).