Faculty Bibliography

OBJECTIVES/OBJECTIVE:The gut microbiome plays a crucial role in regulating systemic immunity and has been implicated in several chronic inflammatory diseases. Intestinal expansions of Ruminococcus gnavus (RG), a dominant gut commensal, correlate with disease flares in lupus nephritis (LN), but the underlying mechanism remains unknown. METHODS:In a Pilot cohort of patients with biopsy-proven LN, subsetted by gut microbiota community, immune status was characterised using bulk-blood RNA sequencing libraries, serum levels of representative host proteins, and levels of immunoglobulin (Ig)G antibodies to the novel lipoglycan (LG) produced by pathogenic RG strains. A Validation LN cohort was evaluated for blood transcriptomic profiles and levels of anti-LG antibodies. In murine models, mechanistic hypotheses were tested after RG gut colonisation or after intraperitoneal injection with an LG preparation, with outcomes determined by transcriptomic analyses, platelet functional readouts, and tissue histology. RESULTS:In a Pilot cohort of patients with LN, RG gut expansions were associated with high-level platelet, neutrophil, and monocyte activation. Serum levels of platelet factor 4 and release of neutrophil extracellular traps (NETs) were significantly higher in patients with high serum IgG antibody against the novel RG-specific LG, a marker of in vivo immune exposure. An LN Validation cohort confirmed these correlates and showed that anti-LG antibodies serve as a surrogate for thromboinflammatory profile in this LN-associated endotype. In mice, gut colonisation with LG-producing RG strains or a single LG injection caused megakaryocytosis and platelet activation; RG colonisation with LG-producing strains induced tubulointerstitial injury with NETosis. In vivo responses to LG toxin were Toll-like receptor 2-dependent. CONCLUSIONS:Gut expansions of the RG pathobiont may contribute to autoimmune pathogenesis through the LG toxin and cause LN flares through thromboinflammatory mechanisms in this previously unrecognised LN endotype.

BACKGROUND:Nondiagnostic cytology for thyroid nodules, consistent with The Bethesda System for Reporting Thyroid Cytopathology category I (B1) poses a management dilemma for clinicians. The objective of this study was to define the malignancy risk of nodules with B1 cytology using American College of Radiology Thyroid Imaging Reporting & Data System (TI-RADS) and to assess whether TI-RADS can help guide the decision to perform a repeat biopsy of these nodules. MATERIALS AND METHODS:This retrospective cohort study evaluated 139 B1 nodules that had a definitive diagnosis on repeat biopsy or surgical excision. Sonographic features were evaluated and classified according to TI-RADS. TI-RADS category and total points were compared to the final diagnosis to determine the malignancy risk of B1 thyroid nodules. RESULTS:Of the 139 nodules, 11 (7.9%) were malignant. The malignancy risk of nodules assigned TI-RADS category 1 and 2 were both 0%, TI-RADS 3 was 2.9%, whereas TI-RADS 4 and 5 were 5.9% and 46.2%, respectively. The optimal cutoff for TI-RADS points predicting malignancy was 5 points. CONCLUSION:B1 thyroid nodules in TI-RADS categories 1-3 may not require repeat biopsy given low malignancy risk. However, B1 nodules in TI-RADS categories 4 and 5 have a higher malignancy risk and thus should undergo repeat biopsy.

OBJECTIVE:This study was undertaken to develop and validate a deep survival model (EEGSurvNet) that analyzes routine electroencephalography (EEG) to predict individual seizure risk over time, comparing its performance to traditional clinical predictors such as interictal epileptiform discharges (IEDs). METHODS:We conducted a retrospective cohort study including 1014 consecutive routine EEGs from 994 patients recorded at a tertiary epilepsy center. We developed EEGSurvNet, a deep learning model that predicts time to next seizure over a 2-year horizon from a single EEG. Model performance was evaluated on a temporally shifted testing set of 135 EEGs from 115 patients using time-dependent area under the receiver operating characteristic curve (AUROC), AUROC integrated over 2 years (iAUROC), and C-index. We compared the deep survival model to a clinical Cox model incorporating standard risk factors as well as a random model based on baseline seizure risk. RESULTS:EEGSurvNet achieved a 2-year iAUROC of .69 (95% confidence interval [CI] = .64-.73) and C-index of .66 (95% CI = .60-.73), outperforming both clinical and random models. Performance was highest in the first months following EEG, peaking at 2 months (AUROC = .80). Combining EEGSurvNet with clinical predictors further improved performances (iAUROC = .70, C = .69). Notably, the model showed superior discrimination on EEGs without IEDs (iAUROC = .78 vs. .53). Model interpretation revealed that the temporal-occipital regions and 6-15-Hz frequencies contributed most to risk prediction. SIGNIFICANCE/CONCLUSIONS:EEGSurvNet demonstrates that deep learning can extract prognostic information from routine EEG beyond visible epileptiform abnormalities, potentially improving patient counseling and treatment decisions. Future prospective studies are needed to validate these findings and assess their clinical impact.

In presynaptic nerve terminals, the endocytic apparatus rapidly restores synaptic vesicles after neurotransmitter release. Many endocytic proteins localize to the periactive zone, a loosely defined area adjacent to active zones. A prevailing model posits that recruitment of these endocytic proteins to the periactive zone is activity-dependent. We here show that periactive zone targeting of endocytic proteins is largely independent of active zone machinery and synaptic activity. At mouse hippocampal synapses and Drosophila neuromuscular junctions, pharmacological or genetic silencing resulted in unchanged or increased levels of endocytic proteins including Dynamin, Amphiphysin, Nervous Wreck, Endophilin A, Dap160/Intersectin, PIPK1γ and AP-180. Similarly, disruption of active zone assembly via genetic ablation of active zone scaffolds at each synapse did not impair the localization of endocytic proteins. Overall, our work indicates that endocytic proteins are constitutively deployed to the periactive zone and supports the existence of independent assembly pathways for active zones and periactive zones.

In recent years, the Permo-Triassic rocks of Kashmir have attracted considerable attention due to their intriguing preservation of fossil assemblages and relatively continuous sedimentary sequences, offering key insights into the end-Permian mass extinction event and subsequent biotic recovery. Among these, the Mandakpal section on the southern limb of the Vihi Syncline has remained comparatively understudied, despite its rich fossil assemblages and apparently continuous Permo-Triassic sequence, similar to the well-studied Guryul Ravine. In this study, the Mandakpal section is re-investigated through detailed fieldwork, lithological characterization, and fossil sampling. Multiple fossiliferous horizons were identified, including beds rich in productid brachiopods and pectinid bivalves. Notably, a partially preserved Meckel's cartilage of a chondrichthyan fish, possibly representing an eugeneodontiform, is documented from the Upper Permian limestone and shale beds of the Zewan Formation. This represents the first non-dental chondrichthyan fossil record from the Permian of Kashmir, highlighting the exceptional preservation potential of these rocks. In addition, the occurrence of Ophiceras ammonoid beds and the bivalve Claraia in the overlying strata confirms the presence of the earliest Triassic (Scythian) at Mandakpal, marking a clear Permo-Triassic transition within the section. The Mandakpal section represents an important locality for regional biostratigraphic correlation, documenting vertebrate fossils, and can contribute to future research aimed at more precise demarcation of the Permian-Triassic boundary in the Tethyan belt, thereby refining our understanding of biotic turn over and environmental disruptions during earth's most severe extinction event.

OBJECTIVES:Increased cancer survival has led to a growing population of geriatric patients with cancer. The post-acute rehabilitation outcomes of these patients are poorly understood. This study aimed to investigate whether cancer status is associated with functional outcomes, new institutionalization, and mortality in Australian geriatric rehabilitation patients. DESIGN:Observational longitudinal REStORing health of acutely unwell adulTs (RESORT) cohort. SETTING AND PARTICIPANTS:Geriatric rehabilitation patients in a tertiary hospital in Australia. METHODS:Patients were allocated to "no," "past," and "active" cancer status groups. Function was measured at geriatric rehabilitation admission and discharge using Katz Activities of Daily Living (ADL) and Lawton-Brody Instrumental ADL (IADL) scores. Logistic regression and Cox regression were performed to investigate the association between functional change with new institutionalization and 1-year mortality, respectively. RESULTS:Of 1890 patients, 77.7% had "no," 13.3% had "past," and 9.0% had "active" cancer. ADL and IADL scores improved in all groups, with no significant difference between groups in functional outcomes or new institutionalization. Compared with "no" cancer (2.9%), a significantly higher proportion of "active" (8.2%) and "past" (6.0%) cancer patients died during the hospital admission, P < .001. Active cancer was associated with higher 3- and 12-month mortality (20% and 40.6%). ADL and IADL improvement were significantly associated with reduced risk of 12-month mortality in "no" and "active" groups, and lower incidence of institutionalization in all groups. CONCLUSIONS AND IMPLICATIONS:Although inpatient mortality is higher in active cancer patients, older adults with active cancer who survive until inpatient geriatric rehabilitation discharge achieve similar functional gains and rates of discharge to the community. Functional improvement is of prognostic value in active cancer patients.

We present ReXVQA, the largest and most comprehensive benchmark for visual question answering (VQA) in chest radiology, comprising 694,841 questions paired with 160,000 chest X-rays studies across training, validation, and test sets. Unlike prior efforts that rely heavily on template based queries, ReXVQA introduces a diverse and clinically authentic task suite reflecting five core radiological reasoning skills: presence assessment, location analysis, negation detection, differential diagnosis, and geometric reasoning. We evaluate eight state-of-the-art multimodal large language models, including MedGemma-4Bit, Qwen2.5-VL, Janus-Pro-7B, and Eagle2-9B. The best-performing model (MedGemma) achieves 83.24% overall accuracy. To bridge the gap between AI performance and clinical expertise, we conducted a comprehensive human reader study involving 3 senior radiology residents on 200 randomly sampled cases. Our evaluation demonstrates that MedGemma achieved superior performance (83.84% accuracy) compared to human readers (best radiology resident: 77.27%), representing a significant milestone where AI performance exceeds human evaluation on chest X-ray interpretation. The reader study reveals distinct performance patterns between AI models and radiology residents, with strong inter-reader agreement among the human readers while showing more variable agreement patterns between human readers and AI models. ReXVQA establishes a new standard for evaluating generalist radiological AI systems, offering public leaderboards, fine-grained evaluation splits, structured explanations, and category-level breakdowns. This benchmark lays the foundation for next-generation AI systems capable of mimicking expert-level clinical reasoning beyond narrow pathology classification.Dataset: https://hf.co/datasets/rajpurkarlab/ReXVQA; Supplementary Material: https://github.com/monk1337/PSB-Conference-2025/blob/main/Appendix.pdf.