Sage Bionetworks

Multi-omics in combination with advanced computational methodologies synthesizes diverse omics data to provide deeper insights into molecular interactions and offers transformative potential for unravelling phenomenon behind disease complexities, improving diagnostics, disease prevention, and personalized treatments. This integrative strategy enables our understanding of gene-environment relationships, chronic disease progression, and the intricate molecular pathways involved in health. Effective multi-omics analyses require robust data sharing, accessibility, interoperability, and governance, which are critical for linking genomic elements to phenotypic traits. The Global Alliance for Genomics and Health advocates for responsible data-sharing practices, by promoting key principles such as transparency and equity. By emphasizing a collaborative approach to data utilization, our proposed framework seeks to advance improved disease prevention and treatment strategies. Multi-disciplinary collaboration, encompassing researchers, clinicians, policy makers, and patient representatives, is pivotal for driving innovation and addressing rare disease diagnostics. The success of multi-omics applications hinges on the establishment of comprehensive datasets, understanding the functional implications of multi-omic variation, adherence to findable, accessible, interoperable, reusable (FAIR) and Collective Benefit, Authority to Control, Responsibility, and Ethics (CARE) principles, and the strengthening of global genomic commons, benefiting scientific research, drug development, and broader health initiatives. Our review highlights essential components of multi-omics integration, underscoring its potential to transform the landscape of precision medicine and improved patient outcomes worldwide.

Abstract Disparities in formal bioinformatics training exacerbate the global skills gap, impeding the democratized application of advanced genomic technologies. To bridge this divide, we introduce a scalable, hybrid training framework designed to rapidly accelerate regional bioinformatics capacity. We exemplify this approach through the Eastern European Bioinformatics and Genomics (EEBG) workshop series — a cross-disciplinary initiative that pairs international faculty with local institutions to deliver modular, hands-on curricula. Functioning as a structured knowledge-transfer pipeline, the series has catalyzed a sustainable educational ecosystem, evidenced by the establishment of multiple independent summer schools across the region. The assessment of the 2025 EEBG workshop in Kraków, Poland, validates the model’s viability; participant metrics confirm high efficacy in skill acquisition (mean satisfaction: 4.4/5.0) and community building. Crucially, the hybrid delivery mode dismantled geographic barriers, serving as a vital mechanism for maintaining scientific continuity for researchers facing displacement and crisis. Synthesizing these outcomes, we define the core features of a replicable blueprint for scientific readiness in resource-constrained environments. We conclude by presenting a strategic roadmap — organized around infrastructure standardization, governance sustainability, and geographical expansion — for adapting this regional proof-of-concept into a global export-ready model, offering a critical path toward ensuring universal access to genomic innovation.

When evaluations aren't trustworthy, entire research programs can chase mirages. Objective benchmarks and independent assessment have repeatedly catalyzed progress across computational biology, from protein structure prediction to variant interpretation and single‑cell analysis. This workshop gathers leaders of community challenges and benchmarking infrastructures together with domain experts to provide a contemporary view of how to design trustworthy evaluations, why blind prediction matters, and where standards, infrastructure, and policy must evolve to meet the demands of AI‑driven biology. We summarize the motivation and scope of the workshop; provide background on methodological and infrastructural advances that enable rigorous benchmarking; highlight invited speakers' contributions; and outline anticipated outcomes and community calls‑to‑action.

INTRODUCTIONSynaptic spine loss in Alzheimer's disease (AD) contributes to cognitive decline. p21‐activated kinase 1 (PAK1), a regulator of spine integrity, is aberrantly activated in AD. We investigated whether PAK1 inhibition might preserve dendritic spines in vitro and in vivo.METHODSOligomeric amyloid beta (oAβ) or tau (oTau) were applied to hippocampal neurons ± NVS‐PAK1‐1, a selective PAK1 inhibitor. NVS‐PAK1‐1 was orally administered to 5xFAD mice. The effects of NVS‐PAK1‐1 treatment on PAK1 activity, spine density, and the proteome were assessed using phospho‐PAK1 (pPAK1) western blotting, Golgi staining, and mass spectrometry for proteomic analyses.RESULTSNVS‐PAK1‐1 prevented oAβ and oTau–induced spine loss in vitro. In 5xFAD mice, NVS‐PAK1‐1 demonstrated brain exposure after oral administration and reduced PAK1 activation, prevented spine loss, and partially normalized synaptic proteomic signatures in females in absence of alterations in brain or plasma Aβ.DISCUSSIONPAK1 inhibition enhances spine resilience in AD models, supporting its therapeutic potential.Highlightsp21‐activated kinase 1 (PAK1) inhibitors prevent oligomeric amyloid beta (oAβ) and oligomeric tau–induced spine loss and dendritic degeneration in cultured mouse hippocampal neurons.NVS‐PAK1‐1, a selective PAK1 inhibitor, protects against oAβ‐induced spine loss in a dose‐dependent manner (EC50 = 2 nM).Oral administration of NVS‐PAK1‐1 achieves brain penetration and bioavailability in normal CD‐1 mice, and target engagement in 5xFAD mice.Chronic NVS‐PAK1‐1 treatment mitigates spine loss in the somatosensory cortex of 6‐month‐old 5xFAD female mice.Chronic treatment with NVS‐PAK1‐1 restores proteomic abundance of actin cytoskeleton and dendritic spine‐associated proteins, including cofilin 2 and pyruvate dehydrogenase kinases, downstream of PAK1 in young 5xFAD female mice showing spine resilience.Clinical oncology trials with other PAK1 inhibitors support potential repurposing or novel compound development for Alzheimer's disease trials.

BackgroundAlzheimer's disease has differing effects on brain regions and cell populations as it progresses. Examining vulnerable cell populations can be difficult at high resolution, however, because large‐scale single cell transcriptomics is expensive, and differential survivability of cells can affect population distribution and quality. Bulk RNA sequencing allows for larger sample sizes and includes nuclear, cytoplasmic, and distal RNA, but does not provide cell‐specific data. In this study, we leveraged the benefits of both modalities by using existing single cell RNA data to deconvolve cell type proportions in bulk RNA samples of human brains from three large‐scale AD studies. We used the best outputs from multiple deconvolution strategies to estimate cell population changes associated with Alzheimer's disease.MethodWe analyzed three bulk RNA data sets from donors with and without AD, provided by the Mayo RNAseq Study (n = 481), the Mount Sinai Brain Bank Study (n = 974), and the Religious Orders Study/Memory and Aging Project (ROSMAP) (n = 1598). Gene marker lists for each cell type were created using five AD‐related single‐cell RNA data sets (Cain 2020, Gabitto 2024, Lau 2020, Leng 2021, and Mathys 2019). We used six deconvolution algorithms (CIBERSORTx, DeconRNASeq, Dtangle, DWLS, MuSiC, and Scaden) with varying marker sets to estimate the cell type proportions in each bulk sample. We also tested different normalization and batch correction strategies. The three best‐fitting parameter sets from each algorithm were used to calculate cell type proportion differences between AD and control samples.ResultWe found significant AD‐related decreases (p < 0.01) in 5 of the 17 neuronal subtypes we analyzed, replicated across multiple algorithms and brain tissues. We also found significant increases in astrocytes and oligodendrocytes. Methodologically, we found that CIBERSORTx, dtangle, and MuSiC, combined with markers generated from the Mathys 2019 data, produced the best results.ConclusionOur research identified five neuronal subtypes that may be more vulnerable to Alzheimer's disease. This knowledge can motivate more experiments at single‐cell resolution that focus on these populations. Our experimentation with different deconvolution methods also yielded useful information for future analysis on similar data with closely‐related or rare cell types.

Metabolic dysregulation is a hallmark of Alzheimer’s disease (AD), with numerous studies characterizing metabolic pathways associated with AD onset and progression. A significant limitation of these studies has been a predominant focus on non-Hispanic white participants. Despite evidence that AD prevalence, progression, and biomarkers differ across ethnoracial groups, it remains unclear whether previously identified metabolic dysregulation in AD brains generalizes across populations. We addressed this gap by analyzing large-scale metabolomics data from 547 postmortem dorsolateral prefrontal cortex brain tissue samples of Hispanic American, Non-Hispanic African American, and White subjects, providing the largest multiethnic AD brain cohort analyzed to date. A metabolome-wide association study examined how relationships between metabolite abundance and AD neuropathology varied by ethnoracial group. Sixty metabolites exhibited significant heterogeneity with tau pathology (Braak stage), with enrichment in tricarboxylic-acid-cycle intermediates, dipeptides, and sphingolipid-ceramide pathway lipids. These findings reveal ethnoracial-specific metabolic signatures of tau pathology and emphasize the need to evaluate emerging therapeutic targets across diverse groups.

BackgroundFalls and cognitive impairment are major sources of disability in Parkinson disease (PD). The ability to accurately identify individuals with PD at high risk for falls and cognitive impairment would provide an opportunity for intervention and potentially improve long-term outcomes. In a previous study, Assessing Telehealth Outcomes in Multiyear Extensions of Parkinson Disease Trials (AT-HOME PD), we remotely characterized participants with early PD who had participated in 1 of 2 PD clinical trials over 2 years of follow-up. These participants with advancing disease provide a unique opportunity to examine whether the capture of objective in-home measures via digital tools and bothersome symptoms via direct participant report improves the prediction of disease milestones.ObjectiveAssessing Telehealth Outcomes in Multiyear Extensions of Parkinson Disease Trials–2 (AT-HOME PD2) aims to examine whether digital tools and remote participant reporting can improve the prediction of falls and cognitive impairment, quantify changes in physical activity over time, and explore the relationship between physical activity and clinical progression over time.MethodsThis is a decentralized observational study of up to 200 individuals with PD, with clinical and digital phenotyping for up to 3 years of follow-up. Participants are those who took part in the STEADY-PD III (NCT02168842), Study of Urate Elevation in Parkinson’s Disease, Phase 3 (SURE-PD3; NCT02642393), AT-HOME PD (NCT03538262), or PD GENEration (NCT04057794) studies. All participants complete 2 video visits per year, wear 2 wrist-worn sensors (Fitbit Charge 5 and ActiGraph CentrePoint Insight Watch) for 1 week each month, complete smartphone-based motor tasks (using the mPower 2.0 app) for 10 days each quarter, and complete online surveys (within the companion Fox Insight study) each quarter. Falls are assessed via a weekly automated telephone call. A cognitive diagnosis is determined by a consensus committee that considers scores on a global cognitive measure, detailed neuropsychological tests, a cognitive-related disability measure, and clinical information. Prediction models will be constructed, and prediction accuracy will be compared across the models.ResultsRecruitment for the study was initiated in September 2023. Enrollment is ongoing, with 142 participants enrolled as of January 2025. Within the cohort, the average age is 69.2 (SD 8.7) years; 85 (59.9%) participants are male, 137 (96.5%) are White, and 2 (1.4%) are Hispanic or Latino; and the average disease duration is 8.9 (SD 1.3) years.ConclusionsAT-HOME PD2 is remotely clinically and digitally phenotyping participants with midstage PD to predict falls and cognitive impairment and to provide insights into long-term progression.International Registered Report Identifier (IRRID)DERR1-10.2196/71955

Abstract The Cancer Complexity Knowledge Portal (CCKP) is an open-access data repository that stores and shares resources from multiple cancer research programs. Maintained by Sage Bionetworks, in collaboration with The Multi-Consortia Coordinating (MC2) Center, which supports six NCI Division of Cancer Biology (DCB) funded programs made up of interdisciplinary communities of scientists, the CCKP centralizes access to cancer research tools and data, improving their discoverability and reuse.Reusing tools has proven to be challenging for various reasons. Therefore, we aim to enhance the reusability of the tools available on CCKP. To achieve this, we are designing and implementing a Cancer Complexity Toolkit to validate tool reusability and increase their adoption. Building on existing efforts to curate tools generated by the cancer research community, the 'Cancer Complexity Toolkit' aims to achieve three key objectives: (1) enhancing the discovery and reuse of cancer research tools through rich metadata and standardized evaluation frameworks, (2) creating a catalog of cancer research-focused tools to improve accessibility and confidence in their use, and (3) enabling standardized reporting and validation of tool functionality. Ultimately, we aim to provide a robust platform for discovering, validating, and integrating computational tools in cancer research.To achieve these goals, we are (1) refining a standardized schema to capture metadata on tool parameters and usage, (2) building infrastructure for automated checks to assess documentation, test coverage, and community metrics, and (3) developing tool cards to display relevant metrics and metadata, helping researchers select the right tools and allowing developers to highlight their tools' impact.In Phase 1, we refined the tool metadata schema and developed a Python-based testing infrastructure to check for documentation, testing files, and deployment on non-native Linux environments. We selected ten tools from the CCKP for initial testing, generating logs that assess their reusability. Citation Format: Aditi Gopalan, Brad Macdonald, James Eddy, Thomas Yu, Susheel Varma, Jineta Banerjee, Jineta Banerjee. Enhancing reusability and adoption of cancer research tools: Development of the Cancer Complexity Toolkit [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 1078.

Abstract Applying artificial intelligence (AI) and machine learning (ML) techniques to biomedical problems is predicated on access to clean, high-quality data and reusable software tools. Data and tool repositories offer a vital infrastructure for centralizing, storing, and sharing such resources with the global research community of cross-disciplinary scientists. The Cancer Complexity Knowledge Portal (CCKP) is a NIH-listed domain-specific repository that makes oncology data findable and accessible. The CCKP is developed and maintained by the Multi-Consortia Coordinating (MC2) Center, led by Sage Bionetworks. The MC2 Center supports resource coordination among six cancer-focused interdisciplinary research consortia funded by the National Cancer Institute (NCI). Through this work, the CCKP has emerged as a significant repository for oncology data and tool sharing. To define the metadata standards for sharing resources on the CCKP, we host data models for various data modalities including genomics and imaging, and are developing new models for emerging multimodal data types like spatial transcriptomics. The data models undergo iterative development based on emerging community needs, with versioned releases maintained in a public GitHub repository. These data models power established data management tools developed by Sage Bionetworks, including the Schematic Python package and Data Curator App, to support annotation of data according to FAIR standards. The data models are designed to help researchers link research outputs (e.g., publications, datasets), thus helping the CCKP synthesize and highlight the activities and outputs of the NCI-funded cancer research programs in a meaningful way. The CCKP provides search and filtering capability to accelerate discovery and collaboration in the cancer research community. As of November 2024, the portal hosts curated information on 3, 786 publications, 904 datasets, and 292 computational tools emerging from over 140 research grants. Our data models also incorporate elements of the Cancer Research Data Commons (CRDC) Data Hub data model to support integration and reuse of datasets within the CRDC ecosystem. The recent advances in computational oncology showcase the strength of secondary use of research data and highlight an unmet need to ensure that data and tools available through repositories are reliable, and reusable. We are engaging with scientists, clinicians, and patient advocates, and leveraging user-centered design approaches and structured data models to make cancer data and tools more findable, accessible, and reusable. These ongoing improvements of the portal aim to increase its usability, bringing experimental labs generating data and computational labs using data closer together, to fuel the iterative cycle of scientific discovery. Citation Format: Orion Banks, Ashley Clayton, Aditi Gopalan, Amber Nelson, Stockard Simon, Verena Chung, Amy Heiser, Jay Hodgson, Aditya Nath, Adam Hindman, Milen Nikolov, Adam Taylor, James Eddy, Susheel Varma, Jineta Banerjee. Cancer Complexity Knowledge Portal: a centralized web portal for finding cancer related data, software tools, and other resources [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 1062.

BackgroundThe generalizability of clinical research hinges on robust study designs, which include the recruitment and maintenance of a representative study population. This study examines the evolution of the demographic characteristics of 329,038 participants who enrolled and participated in The All of Us Research Program (AoURP), a decentralized study aimed at representing the diversity of the United States.ObjectiveThe primary objectives of this study were to assess alterations in the demographic composition of the cohort at different protocol stages within AoURP, while analyzing completion rates and timeframes for survey and substudy completion. Additionally, we examined how participant interactions with the program impacted engagement and survey responses.MethodsWe conducted a longitudinal analysis of the AoURP data, tracking changes in demographic composition, completion rates, and completion times for surveys and substudies. Comparative analyses were performed to assess differences in engagement and survey completion based on sociodemographic characteristics of participants involved in postenrollment study components.ResultsThe sociodemographic composition of the cohort that participated in the postenrollment study (eg, optional components) differed significantly from that of the recruited population. The proportion of self-identified White participants increased by 21.2%, whereas the proportion of Black or African American participants decreased by 12.18% (P=.02). Participants who identified as White (n=93,614, 52.7%) and NonHispanic (n=109,279, 42.21%) were more engaged compared to those identifying as Black or African American (n=10,887, 15.76%), Asian (n=4274, 38.72%), or Hispanic (n=12,530, 20.7%; P=.006). Participants’ response times to study surveys and completeness varied across all demographic groups (P<.001). Furthermore, those identifying as White skipped fewer survey questions (1.19) compared to those identifying as Black or African American (1.40) or other racial and ethnic identities (P<.001).ConclusionsThe AoURP dataset serves as an exceptional resource for investigating diverse public health concerns. However, the longitudinal analysis of participant-level data underscores a significant skew in population diversity, suggesting the need for targeted strategies to enhance engagement and retention across all groups. Ensuring diversity in the cohort is essential for maintaining the study’s representativeness and the broad applicability of its findings.