Foundation Models (FMs) are gaining increasing attention in the biomedical artificial intelligence (AI) ecosystem due to their ability to represent and contextualize multimodal biomedical data. These capabilities make FMs a valuable tool for a variety of tasks, including biomedical reasoning, hypothesis generation, and interpreting complex imaging data. In this review paper, we address the unique challenges associated with establishing an ethical and trustworthy biomedical AI ecosystem, with a particular focus on the development of FMs and their downstream applications. We explore strategies that can be implemented throughout the biomedical AI pipeline to effectively tackle these challenges, ensuring that these FMs are translated responsibly into clinical and translational settings. Additionally, we emphasize the importance of key stewardship and co-design principles that not only ensure robust regulation but also guarantee that the interests of all stakeholders—especially those involved in or affected by these clinical and translational applications—are adequately represented. We aim to empower the biomedical AI community to harness these models responsibly and effectively. As we navigate this exciting frontier, our collective commitment to ethical stewardship, co-design, and responsible translation will be instrumental in ensuring that the evolution of FMs truly enhances patient care and medical decision-making, ultimately leading to a more equitable and trustworthy biomedical AI ecosystem.

Background: Stress resilience is a dynamic process shaped by the interaction between demands and adaptive resources. Existing measures assess stress and resilience as separate constructs, limiting their use in digital health and workplace interventions. An integrated measure capturing both domains is needed. Objective: We developed and validated the WONE Index, a multidimensional stress resilience tool designed to measure both current stress load and adaptive resources among full-time working adults. Methods: We developed the 32-item WONE Index through literature review, expert consultation, and iterative refinement to assess stress load and resilience resources across behavioral, cognitive, and social domains. Phase 1 (N=1005; United States– or United Kingdom–based full-time employees) evaluated the initial item pool using exploratory and confirmatory factor analyses to establish the preliminary factor structure and assess reliability and validity. Phase 2 (N=306; United States–based adults) expanded underperforming domains, refined items, and tested incremental validity, test-retest reliability, and measurement invariance. Data were collected online through CloudResearch (Connect) and Prolific (Prolific Academic Ltd) using secure survey platforms. Results: Phase 1 supported a 2-domain structure: a Stress Load factor (Work Stress, Personal Stress, and Burnout) and a Resilience Resources factor (Emotion Regulation and Coping, Social Connectedness, and Sleep). Model fit indices were excellent (comparative fit index, CFI=0.95; Tucker-Lewis index, TLI=0.94; and root mean square error of approximation, RMSEA=0.05). Phase 2 replicated and extended this structure, expanding Resilience Resources into 7 domains (adding Purpose and Prosociality, Physical Activity, Dietary Intake, and Perseverative Thinking). Confirmatory factor analyses supported a 2-domain structure, comprising a higher-order Stress Load factor with 3 subdomains (Work Stress, Personal Stress, and Burnout) and a higher-order Resilience Resources factor with 7 subdomains (Emotion Regulation and Coping, Social Connectedness, Purpose and Prosociality, Sleep, Physical Activity, Dietary Intake, and Perseverative Thinking). The Stress Load model demonstrated excellent fit ( χ ² ₃₃ =64.18; P =.01; CFI=0.99; TLI=0.98; RMSEA=0.06; and standardized root mean square residual=0.05), and the Resilience Resources model also fit well ( χ ² ₄₄₃ =745.20, P <.001; CFI=0.94; TLI=0.94; RMSEA=0.05; and standardized root mean square residual=0.06). All subscales showed strong internal consistency (composite reliability: mean 0.84, SD 0.10; range 0.74‐0.93) and excellent test-retest reliability over 3 weeks (intraclass correlation coefficients 0.77‐0.90, 95% CI 0.87-0.93). The index showed strong convergent validity ( r =0.73 with Connor-Davidson Resilience Scale and r =–0.66 with Perceived Stress Scale-4) and explained additional variance beyond established measures in predicting depression, anxiety, and well-being (Δ R ²=0.07‐0.11; P <.001). Conclusions: The WONE Index provides a psychometrically robust tool for assessing stress resilience capacity in working adults. Its integrated structure captures dynamic relationships between stress exposure and resilience resources, thereby supporting personalized intervention delivery in digital health platforms and organizational well-being programs.

Abstract 391 The mammalian Target of Rapamycin (mTor) is an evolutionary-conserved serine/threonine protein kinase that plays critical roles in energy homeostasis, and cell-decision processes, such as cell growth and proliferation. Genetic alterations that lead to increased mTor activity in humans and mouse models have suggested a critical role for this kinase and its signal-transduction pathway in hematopoietic-stem cell (HSC) biology as well leukemogenesis, and has garnered much attention as a therapeutic target in human hematopoietic diseases. Further confirmation for a critical function of mTor in hematopoiesis has been suggested through pharmacological inhibition with rapamycin. However, recent evidence has shown that rapamycin is an incomplete inhibitor of mTor and its two associated complexes (mTorc1 and mTorc2). We thus sought to definitively discern the role of mTorc1 in hematopoiesis and leukemogenesis by utilizing mice containing inducible “knock-out” alleles of the essential mTorc1 component Raptor. To conditionally ablate the Raptor gene, mice with flanking loxP sites (floxed) of exon 6 of raptor were crossed with mice expressing Cre under the interferon-inducible Mx promoter (MxCre), and deletion induced in 4–6 week old offspring with injections of polyIpolyC. Induced MxCre:raptor homozoygous mice (KO) and controls (WT) were then analyzed for hematopoietic pertubations. Deletion of raptor leads to a complete ablation of mTorc1 biochemical activity in HSC and progenitors (HSPC) as assessed by phosphoflow analysis of key mTorc1 substrates in response to stem-cell factor. Furthermore, cell-type specific pertubations of a major negative feedback loop from mTor to the Akt pathway were uncovered by this analysis, with increased pAkt levels apparent only in early monocytic cells. Loss of mTorc1 activity produced a non-lethal hematopoietic disorder characterized by leucopenia, transient hypochromic anemia, and thrombocytopenia. Bone marrow (BM) of KO mice was characterized by an apparent lack of normal hematopoiesis and a striking expansion of monoyctic cells at the expense of neutrophils and B lymphocytes. These data suggest a critical role for mTorc1 in the execution of myelomonocytic development. Splenomegaly was prevalent in KO animals, reflecting extramedullary hematopoiesis. As many of these hematopoietic alterations could result from defective HSPC function, we examined the HSPC compartment in more detail. Raptor loss leads to an accumulation of mostly G₀ immunophenotypically-defined HSC and multi-potent progenitors in the BM. Metabolic and transcriptional profiling of HSC/MPP indicated pertubations in major metabolic pathways in the KO cells, including decreases in cholesterol and steroid biosynthesis, and alterations in transamination reactions, among others. Strikingly, loss of mTorc1 in HSC/MPP leads to alterations in expression of genes that play critical roles in HSC. To examine the relevancy of these findings to HSC function we performed a series of competitive transplantation experiments. Results from these experiments indicate that mTorc1 is required in a cell-autonomous manner for HSC fitness and proper myelomonocytic differentiation, likely through the control of key metabolic and gene-expression programs. While mTorc1/Raptor loss resulted in profound hematopoietic dysfunction, KO mice survived for greater than 1 year after induced deletion of the raptor gene and maintained hematopoiesis in the absence of Raptor. We next asked whether loss of mTorc1 could be a reasonable therapeutic target in leukemia by generating mice with compound floxed alleles of both Raptor and the Pten gene (deleted again with Cre under the Mx promoter). While homozygous Pten deletion results in a rapidly-fatal myeloproliferative disease in both WT and induced Raptor heterozygous mice, homozygous deletion of Raptor leads to a significantly prolonged survival. Furthermore, through transplantation experiments, we show that mTorc1 is required in a cell-autonomous manner for Pten-loss-evoked leukemogenesis. These studies have demonstrated critical functions of mTorc1 in benign hematopoiesis, suggest rapamycin-independent functions of mTorc1 in this system, and imply that mTorc1 ablation can have beneficial effects in mouse models of leukemogenesis.