Abstract Collaborative cancer research requires a shared data infrastructure that functions across institutions, modalities, and funders. Synapse.org, developed by Sage Bionetworks, is a versatile public data platform used by research consortia, real-world data efforts, and computational challenges. We describe the use of Synapse to support diverse cancer research communities. Synapse provides tools for validated metadata, provenance, versioning, and fine-grained access control. For supported communities we provide curation and validation apps, domain-specific portals, dashboards, and challenge frameworks. APIs and clients connect Synapse to local compute and trusted research environments (Cavatica, Terra, Pluto, SevenBridges CGC) via GA4GH DRS. Recent enhancements include human-guided, AI-powered curation, OpenSearch-based discovery, and natural-language data search. Data arrive through contributor uploads (web, CLI, APIs), programmatic ETL pipelines, and indexing of external repositories (GEO, dbGaP).. Curated datasets are indexed for discovery and surfaced through program portals. Community use and feedback guide curation priorities, maintaining a continuous improvement cycle. The Synapse platform hosts >3.6 PB data used by >6,000 monthly users. The Cancer Complexity Knowledge Portal links 160 grants, 4,178 publications, 1,039 datasets, and 321 tools focused on cancer biology. The Human Tumor Atlas Network DCC manages 334 TB of harmonized omics data (0.23M files from 2,372 cases and 11,378 biospecimens across >60 diseases and 25 assays), with >2,500 annual users. The NF Data Portal integrates over 200 TB data from 312 neurology and oncology studies, and catalogues >1,100 tools spanning NF1, NF2, and schwannomatosis. As part of the coordinating center for AACR Project GENIE we have supported the ETL and sharing of clinico-genomic data from 19 centers, including 211,527 patients and 250,018 samples. DREAM cancer challenges run on Synapse have established benchmarks including in AML subtyping, prostate cancer survival, immunotherapy response and digital mammography. Independent Synapse users have contributed >1.2M public files (>95 TB) across 400+ cancer-related projects. A user-friendly generalist data platform and a set of proven data coordination and challenge operations lower the activation energy for new cancer collaborations, from independent projects to large, multi-institutional networks. We invite researchers to explore Synapse.org and our data portals. AI was used to summarize metrics and refine wording Authors reviewed and approved all content. Citation Format: Adam Taylor, Gaia Andreoletti, Robert Allaway, Jineta Banerjee, Orion Banks, Angie Bowen, Kevin Boske, Aditi Gopalan, Ashley Clayton, Xindi Guo, Savitha Sangameswaran, Amber Nelson, Aditya Nath, Milen Nikolov, Anh Nguyet Vu, Ziwei Pan, Alex Paynter, Chelsea Nayan, Thomas Yu, Bishoy Kamel, Serghei Mangul, Alberto Pepe, Luca Foschini, Susheel Varma. Synapse.org as a foundational platform supporting multicenter cancer data coordination, benchmarking, and broad community reuse [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 1.
Background:Advances in medicine depend on analyzing large and complex data sources, but discovery is partly constrained by the limited time and domain expertise of human researchers. Agentic artificial intelligence (agentic AI) can accelerate discovery by automating components of the scientific workflow, including information retrieval, data analysis, and knowledge synthesis. Aim:OpenScientist, an open-source agentic AI co-scientist, aims to accelerate biomedical discovery by semi-autonomously investigating scientist-defined queries and generating clinically relevant, verifiable scientific insights. Methods:Domain experts evaluated OpenScientist for novel discoveries in four clinical case studies: (1) a prespecified analysis in a community-based Alzheimer's disease biomarker cohort, (2) unsupervised modeling for plasma proteomic survival prediction, (3) hypothesis investigation in single-cell transcriptomic data from neurons with neurofibrillary tangles, and (4) hypothesis generation with validation in a multiple myeloma dataset with a randomized negative control. Results:OpenScientist completed analyses in minutes that otherwise would take weeks to months of human time and expertise. It identified %ptau217 as the best predictor of amyloid PET status, generated a plasma proteomic survival model with performance comparable to published models, proposed a mechanism linking tau pathology to altered lysosomal acidification, and generated multiple myeloma hypotheses that were validated in an external cohort while distinguishing true signal from randomized controls. Conclusion:OpenScientist demonstrates that open, auditable, agentic AI can support real-world clinical research by generating hypotheses, executing analyses, and discovering insights from complex datasets.
Neurofibromatosis type 1 (NF1) is an autosomal dominant condition in which patients are heterozygous for a disruptive pathogenic variant in the NF1 gene. The most characteristic feature of the condition NF1 is the neurofibroma, a benign, multi-cellular tumor which initiates when a cell of the Schwann cell lineage gains a somatic pathogenic variant of the other NF1 allele. Neurofibromas developing at nerve termini in the skin are termed "cutaneous" neurofibromas (cNFs), while those developing within larger nerves are termed "plexiform." Most patients develop cNFs beginning in late childhood or early adulthood, continuing throughout life at variable rates. Some patients may develop only a few cNFs, while others suffer from thousands. There are no reliably effective physical or pharmaceutical therapies besides surgical removal. Although these are not life-threatening, they are disfiguring and can interfere with normal life functions. To provide a resource for research, we developed short-term cNF Schwann cell cultures from NF1 patients, from which we subsequently established the first semi-immortalized cNF cell lines through transduction with wild-type human telomerase reverse transcriptase (hTERT) and murine cyclin-dependent kinase 4 (mCdk4) genes. Here we present molecular, cellular, and functional characterization of these cell lines, which will be of utility for investigating and developing NF1 cNF therapies.
Croissant has emerged as the metadata standard for machine learning datasets, providing a structured, JSON-LD-based format that makes dataset discovery, automated ingestion, and reproducible analysis machine-checkable across ML platforms. Adoption has accelerated, and NeurIPS now requires Croissant metadata in every submission to its dataset tracks. Yet in practice Croissant generation usually starts with uploading data to a public platform, a path infeasible for governed and large local repositories that hold much of the high-value data ML increasingly relies on. We release Croissant Baker, a local-first, open-source command-line tool that generates validated Croissant metadata directly from a dataset directory through a modular handler registry. We evaluate Croissant Baker on over 140 datasets, scaling to MIMIC-IV at 886 million rows and 374 Parquet files. On held-out comparisons against producer-authored or standards-derived ground truth, Croissant Baker reaches 97–100% agreement across multiple domains.
Abstract Purpose: The Human Tumor Atlas Network (HTAN) sets out to map the cellular and molecular architecture of human tumors over space and time to advance precision oncology. The HTAN Data Coordinating Center (DCC) underpins this mission by standardizing, integrating, and distributing multimodal data; ensuring legacy and impact. Methods: The DCC developed a scalable cloud ecosystem (Synapse.org, Google BigQuery, custom data portal) for data ingestion, governance, validation, and dissemination. A community-driven process, aligned with NCI standards, produced a consistent metadata schema that ensures interoperability across rich clinical and biospecimen information and all data modalities. The HTAN data portal provides a single landing page for users of HTAN data featuring filter-based search, visualization of imaging and single-cell datasets and detailed documentation. Data are disseminated via a tiered model, with imaging and dbGaP-controlled access sequencing data available from NCI Cancer Research Data Commons General Commons and open-access processed data on Synapse.org. Current work transitions to a modular LinkML data model, adds an AI-assisted curation interface, and includes a streamlined medallion architecture in BigQuery and portal enhancements. Results: The first five years of HTAN (release v7.0) produced 334 TB of multimodal data (0.23M files) from 2,372 cases and 11,378 biospecimens, spanning >60 disease types and >25 assays. HTAN supports >3,400 unique global users a month. The most frequently accessed data includes scRNA-seq (1075 downloading users), multiplex imaging (197), and spatial transcriptomics (419). Analysis of dbGaP requests to-date (N=127) shows broad academic (75%) and industry use with research themes focused on genome instability (115 requests), immune evasion (52), and metastasis (49), often with multi-omic and AI-driven approaches. Conclusions: HTAN has established a globally utilized, harmonized foundation for spatially resolved, multimodal cancer research. This enduring platform for community-driven discovery is built on robust data management and transparent access models. The next phase deepens integration of spatial and clinical data, expanding AI-readiness, and strengthening support for reuse and reproducibility. We invite researchers to engage with this resource. ChatGPT 5, Gemini 2.5 Pro, & Claude Sonnet 4.5 were used to summarize usage metrics, conduct thematic analysis of dbGaP applications, and initial abstract drafting. All content was evaluated and approved by the authors. Citation Format: Adam Taylor, Ashley Clayton, Aditi Gopalan, Milen Nikolov, Thomas Yu, David Gibbs, Yamina Katariya, Dar'ya Pozhidayeva, Ino de Bruijn, Seluck Onur Sumer, Kristen Anton, Jennifer Altreuter, Alex Lash, Ethan Cerami, Nikolaus Schultz, Vesteinn Thorsson. Reading the map: An invitation to the resources of the Human Tumor Atlas Network [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 4169.