IntroductionDepression assessment remains largely dependent on symptom reports and clinician judgment, while objective tools for depressive-spectrum stratification and severity estimation remain limited. Existing digital and multimodal depression-detection studies often focus on binary case-control classification, handle missing modalities incompletely, provide limited calibration assessment, and rarely combine depressive-spectrum classification with continuous symptom-severity estimation. We therefore developed a quality-aware multimodal framework integrating eye tracking, facial behavior, and transcript-derived language for classification across normal control (NC), subthreshold depression (SD), and major depressive disorder (MDD), together with prediction of 17-item Hamilton Depression Rating Scale (HAMD-17) severity.MethodsA total of 186 participants completed a controlled task battery including interview, emotional reading, free viewing with verbal description, fixation, gaze orienting, smooth pursuit, prosaccade, and antisaccade tasks. Eye-tracking, facial-video, and transcript-derived language data were converted into modality-specific features. Baseline-3 combined modality-specific encoders, quality-aware gated fusion, and joint classification-regression learning under a nested repeated-resampling framework with explicit missing-modality handling. Baseline-3+ further incorporated Transformer-based cross-modal interaction and uncertainty-based dynamic task weighting. Performance was evaluated on held-out outer-loop test sets after temperature scaling. Interpretability analyses included gate profiling, selective prediction, SHAP, Integrated Gradients, and counterfactual analysis.ResultsBaseline-3+ showed the most favorable classification and calibration profile, with accuracy, balanced accuracy, and F1-macro approaching 0.90 across both classification routes and lower expected calibration error than Baseline-3. For severity estimation, the improvement was route-dependent and mainly reduced the regression disadvantage observed under the hierarchical route. Misclassification was concentrated near the SD boundary. Interpretability analyses showed stable quality-aware modality reweighting, with facial features providing the dominant signal, complemented by eye tracking and smaller but meaningful language contributions.DiscussionThis framework addresses key limitations of prior binary and incompletely calibrated depression-detection models by jointly supporting depressive-spectrum classification, severity estimation, missing-modality handling, calibrated prediction, and individual-level interpretation. Its most plausible role is to augment clinical assessment, particularly for boundary states such as SD.
Generative AI as interactional infrastructure for meaning-centered care in later life
Generative artificial intelligence (GenAI) and large language models are rapidly entering mental health research and service delivery, yet their dominant use remains symptom-centric, emphasizing screening, classification, triage, and risk detection. For older adults, mental health is often inseparable from existential concerns: loss of social role, disrupted continuity of self, loneliness, diminished dignity, and questions of legacy. This perspective argues that GenAI should not be conceptualized as an autonomous substitute for clinicians, nurses, social workers, or family caregivers. Instead, it may be better understood as an interactional infrastructure for meaning-centered care in later life. Drawing on meaning-centered psychotherapy, dignity therapy, life review, gerotranscendence theory, care ethics, and implementation science, we propose a Sensing-Narrating-Connecting-Governing framework. In this model, multimodal AI systems help detect existential and relational cues, support life-review conversations, co-construct dignity-preserving narratives, connect older adults with human care networks, and operate under explicit safeguards for privacy, hallucination, dependency, crisis escalation, and cultural adaptation. The proposed framework shifts evaluation from model performance alone toward existential well-being, dignity, continuity of self, therapeutic alliance, equity, and workflow integration. We conclude that GenAI may contribute to public mental health only when deployed as a bounded, human-supervised, culturally responsive layer of relational augmentation rather than as a replacement for human presence.
Generative AI for pre-consultation mental health triage in disorders of gut-brain interaction
Disorders of gut-brain interaction (DGBI) are common, disabling, and frequently accompanied by anxiety, depressive symptoms, sleep disturbance, symptom-related fear, and repeated health care use. In routine gastroenterology practice, these problems are often recognized late, after fragmented histories, multiple visits, and avoidable investigations. Recent work on generative artificial intelligence (GenAI) and conversational systems suggests a narrower and more practical clinical use case than autonomous diagnosis: supervised pre-consultation triage. We propose that DGBI is a suitable setting for this approach because triage depends on integrating symptom narratives, prior investigations, alarm features, and psychosocial context rather than on a single test result. A GenAI-enabled intake tool could summarize patient-entered histories, incorporate brief distress screening and symptom diaries, flag possible medical or psychiatric escalation, and help route patients toward standard gastroenterology review, integrated psychogastroenterology, dietetic input, or urgent assessment. Its value would lie in making the first consultation more efficient and more clinically informed, not in replacing specialist judgment. For such systems to be acceptable, five conditions are essential: a narrowly defined triage task, multidomain but proportionate data collection, explicit rules for medical and psychiatric escalation, clinician review before action, and prospective evaluation across workflow, safety, equity, and patient acceptability. DGBI offers a realistic opportunity to develop GenAI tools that are useful precisely because they are constrained, auditable, and embedded in multidisciplinary care.
From digital access to social connectedness: the digital divide, bonding social capital, and depressive symptoms among older adults in China
IntroductionAs population aging and digital transformation continue simultaneously in China, the digital divide among older adults has become an increasingly important social issue. This study examines the associations between multiple dimensions of the digital divide and depressive symptoms among older adults, as well as the potential role of bonding social capital.MethodsDrawing on three waves of data from the China Family Panel Studies (CFPS, 2018–2022), this study employs two-way fixed effects models and mediation analyses to examine the relationships between digital access, digital usage, digital outcomes, and depressive symptoms among older adults. Robustness checks were further conducted using propensity score matching (PSM), sample restriction adjustments, and replacement of the dependent variable.ResultsInternet access was significantly associated with lower levels of depressive symptoms among older adults (p < 0.05). Compared with non-Internet users, entertainment-oriented, instrument-oriented, and mixed Internet use were all significantly associated with lower depressive symptoms (all p < 0.05). Digital outcomes were also negatively associated with depressive symptoms (p < 0.01). Bonding social capital showed significant indirect pathways linking all dimensions of the digital divide and depressive symptoms, with mediating proportions ranging from 5.95% to 26.67%. Period heterogeneity analyses further indicated that the associations remained generally stable before and during the COVID-19 period, although mixed Internet use exhibited a significant structural difference across periods (p = 0.036).DiscussionThe findings suggest that the digital divide is closely associated with the mental well-being of older adults, while bonding social capital constitutes an important social pathway linking digital engagement and psychological health. Policy efforts should move beyond technological access toward broader digital empowerment and the construction of a more inclusive digital society for aging populations.
Experiences and wellbeing of family members and carers, regarding PARCS across Victoria
IntroductionWhen consumers experience mental health crises, carers are often key supporters and are also impacted significantly themselves. The Prevention and Recovery Care (PARC) model offers a community-based residential program for consumers, in times of mental health crisis. The potential of PARC services to engage carers is under-examined. This study addresses two questions: How do carers experience the PARC service; and what are carers’ experiences of their own wellbeing, during and after engagement with PARC?MethodsThis is a mixed-methods convergent study of carer experience, whereby quantitative survey data and qualitative survey and interview data were gathered, analysed concurrently, and integrated to report carers perspectives of PARC services. Carers reported their wellbeing across 4 timepoints (n = 71) and also their experience of PARC services in a Carer Exit Survey (n = 50). An independent sample of six family members, each from a different PARC service, engaged in semi-structured telephone interviews.ResultsFor service experience, carers rated the PARC service as highly satisfactory. Interviewees reported a sense of relief, gratitude, and period of regrouping, while valuing the PARC service and feeling positive about accessing PARC services in the future. Positive experience was defined in contrast with distressing experiences of acute wards; concerns were expressed about limits to timely access of PARC services in future if needed. Regarding carer wellbeing, time 1 levels varied across participants, and all measures showed improvement for carers over time. They reported experience of respite, with confidence to entrust their family member to the team, and learning from PARC service staff ways to cope and interact with their family member.DiscussionCarers considered the PARC service a positive environment for the person to receive treatment and support and also experienced PARC services as supporting their own quality of life and wellbeing. This study contributes evidence about how highly valued these recovery oriented sub acute residential services are for carer service users; however, there is potential to further enhance the engagement of carers in PARC service delivery, including through inclusion of carers in co-design.
Author Correction: Quantification and transcriptome profiling reveal abundant, dynamic and translatable dephospho-CoA-capped RNAs
Nature Biotechnology, Published online: 16 June 2026; doi:10.1038/s41587-026-03215-z
Author Correction: Quantification and transcriptome profiling reveal abundant, dynamic and translatable dephospho-CoA-capped RNAs
Engineered heart muscle passes early clinical milestone
Nature Medicine, Published online: 16 June 2026; doi:10.1038/d41591-026-00031-5
Engineered heart muscle allografts derived from induced pluripotent stem cells show promising early outcomes in patients with treatment-refractory advanced heart failure with reduced left ventricular ejection fraction, in support of further clinical investigation.
Author Correction: Global burden of amphetamine, cannabis, cocaine and opioid use in 204 countries, 1990–2023: a Global Burden of Disease Study
Nature Medicine, Published online: 16 June 2026; doi:10.1038/s41591-026-04513-4
Author Correction: Global burden of amphetamine, cannabis, cocaine and opioid use in 204 countries, 1990–2023: a Global Burden of Disease Study
Red-shifted GRAB acetylcholine sensors for multiplex imaging in vivo
Nature Neuroscience, Published online: 16 June 2026; doi:10.1038/s41593-026-02325-w
The authors develop red fluorescent GRAB acetylcholine (ACh) sensors and highlight rACh1h as a robust tool for multiplex recording together with various green sensors. Using fiber photometry, mesoscopic imaging and two-photon imaging, rACh1h is shown to reliably report ACh dynamics in vivo.
STAT+: Verge Labs’ new AI model solves patient stratification problems for neurology clinical trials
As the saying goes, one man’s trash is another man’s treasure. Or as Verge Labs might put it, one company’s failed clinical trial … is that same company’s new AI benchmarking dataset.
Over a decade ago, Alice Zhang co-founded Verge Genomics with the idea that by looking at the network of genes causing neurodegenerative diseases like Parkinson’s, ALS, or Alzheimer’s, the company would be able to come up with better drugs. The company did target discovery work, for example, coming up with two targets that Eli Lilly nominated to its internal pipeline in 2024. Verge also had its own pipeline, where it was pursuing an ALS drug — that is, until its Phase 1b trial failed last month.
The company published a postmortem explaining what exactly went wrong with the trial, in which a third of the patients dropped out because they could not tolerate the drug. “While the temptation is strong, when a trial doesn’t meet the anticipated end points, to kind of look away and not talk about it, we think there are a lot of learnings that can come — not just for us, but for the field and for ALS broadly — that’s really important to share. That’s not done very often,” Zhang told STAT in an interview.