Science Bulletin

Japanese Encephalitis Virus (JEV) belongs to the Flavivirus family, and the RNA dependent RNA polymerase (RdRp) domain located at the C terminus of non structural protein 5 (NS5) regulates de novo viral genome synthesis. The conserved priming loop inserted in the thumb domain of RDRP initiates de novo genome synthesis. Interestingly, the reported replication process initiates with an interaction between NS5 (methyltransferase/RDRP) and NS3 (protease/helicase), and inhibiting this interaction directly correlates with the inhibition of viral replication. The function of the priming loop in the context of the NS5 NS3 interaction is not yet known. In this study, we studied the priming loop function in the NS5 NS3 interaction and viral replication. Using a structure based drug design approach, we screened the Maybridge compound library against the RdRp priming loop and identified 15 candidate compounds based on binding energy. Among them, DSHS00151 showed significant dose dependent inhibition of the NS5 NS3 interaction with an IC50 of 3.5 micromolar in the mammalian two hybrid assay, inhibited viral infectivity with an IC50 of 2.54 micromolar, and reduced viral RNA load with an IC50 of 2.9 micromolar, compared to CD10712. Molecular Dynamics (MD) simulations revealed that DSHS00151 exhibits stronger and more stable binding with the priming loop residues (790 to 812) compared to CD10712. The molecular mechanism suggested by the docking of the NS5 and NS3 proteins showed that the priming loop tends to rotate toward the NS3 protein to stabilize the complex, and compound binding restricts this loop rotation, thus compromising the stability of the NS5 and NS3 complex and affecting viral replication. Overall, our study validated the function of the NS5 priming loop as an allosteric site, and compounds binding to this allosteric site belong to the Non Nucleoside Reverse Transcriptase Inhibitors (NNRTI) class of inhibitors, which disrupt the NS5 and NS3 interaction and could be developed as novel anti JEV therapeutics.

The long-standing challenge of seasonal influenza vaccines providing poor protection is attributed to the rapid and continuous evolution of the virus. In theory, the recommendation of vaccine strain is to predict the dominant strain with best fitness in the upcoming season. Here, we develop FutureFlu, a biologically-grounded framework that quantifies the fitness score of influenza variants in upcoming seasons, by integrating metrics on three levels: molecular genetic divergence, individual immune escape, and population-scale transmission dynamics. The fitness scores of influenza variants present strong positive correlation with their actual observed frequencies in the next seasons. Validation across 24 seasons of three influenza subtypes shows FutureFlu recommends antigenically matched vaccine strains more frequently than annual recommendations, particularly for challenging subtypes: 83.3% versus 45.8% seasons for A/H3N2, and 75.0% versus 33.3% seasons for B/Victoria. Furthermore, FutureFlu significantly outperforms the currently used methods in vaccine strain selection whether or not there is available antigenic data from hemagglutination inhibition (HI) assay, which provides a valuable supplement for WHO vaccine recommendation. To support global public health implementation, an open online platform (futureflu.com.cn) has been established to offer real-time viral fitness predictions and vaccine recommendations.

Every known life form senses and reacts to mechanical forces. These mechanical stimuli can be converted into electrical signals by mechanically gated ion channels, a transduction cascade pivotal to numerous physiological functions including touch, hearing, mechanical pain, circulation, gastrointestinal function, and mechanical loading in various tissues. Despite continuous efforts, numerous mechanically gated ion channels with the mechanotransduction process underlying these physiological functions remain unidentified. Here, we focused on the transmembrane channel-like (TMC) protein family expressed in the cultured cells to identify those with potential mechanosensitive activity. Remarkably, in contrast to human TMC1/2 (HsTMC1/2), human TMC3-8 (HsTMC3-8) proteins are localized to the plasma membrane when heterologously expressed in the cultured cells. Further experiments revealed that mechanical poking stimuli can effectively activate HsTMC3-8. In addition, HsTMC3-8 induced stretch-activated currents and elicited well-resolved single-channel activities in response to negative pressure stimulation. The mutants near the putative pore region altered reversal potentials (Erev) of HsTMC3-8, suggesting that TMC3-8 are likely pore-forming subunits of ion channels. In summary, we proposed that TMC proteins are the largest mammalian mechanically gated ion channel family.

BACKGROUND: Coronary angiography remains the reference standard for diagnosing coronary artery disease and guiding revascularization, yet its interpretation requires expert integration of multi-view anatomy, lesion morphology and procedural context. Existing artificial intelligence approaches are largely task-specific, annotation-dependent and limited in capturing the semantic relationship between angiographic findings and interventional decision-making. Whether large-scale vision-language pretraining can enable transferable foundation-model representations for invasive coronary imaging remains unknown. METHODS We developed CAG-MIND, a domain-specific vision-language foundation model for coronary angiography, using 135,475 CAG examinations paired with procedural reports, comprising 812,850 angiographic videos from Zhongshan Hospital and Shanghai Geriatric Medical Center. Each case consisted of standardized six-view angiographic acquisitions paired with structured procedural semantics extracted from routine reports using a large language model-assisted pipeline. The model was pretrained by aligning multi-view angiographic representations with report-derived semantic embeddings through bidirectional contrastive learning. Performance was evaluated under zero-shot and supervised fine-tuning settings across 11 downstream tasks grouped into structural abnormality detection, atherosclerotic plaque assessment, and interventional decision prediction, using both an internal validation cohort and an independent external test cohort. RESULTS CAG-MIND demonstrated robust performance across all three task categories. In the zero-shot setting, the model achieved mean AUROCs of 0.686 in the internal validation cohort and 0.745 in the external test cohort, indicating transferable multimodal representations without task-specific supervision. Following supervised fine-tuning, the mean AUROC increased to 0.827 and 0.846, respectively, with excellent performance for coronary stenosis detection (AUROC 0.940 in both cohorts), balloon/stent prediction (0.900 and 0.907), and CABG recommendation (0.877 and 0.875). Compared with representative biomedical vision-language models and conventional image-based architectures, CAG-MIND consistently achieved superior performance in both zero-shot and supervised settings and remained superior to fully fine-tuned competing models when trained with only 10% of the labelled data. Grad-CAM visualization demonstrated anatomically plausible lesion-focused attention, supporting the interpretability of the learned representations. CONCLUSIONS CAG-MIND is, to our knowledge, the first large-scale vision-language foundation model for coronary angiography trained at more than 100,000-patient scale. By aligning standardized multi-view angiographic videos with report-derived procedural semantics, CAG-MIND enables robust zero-shot transfer, data-efficient fine-tuning and cross-center generalization. These findings support domain-aligned multimodal pretraining as a scalable foundation-model paradigm for invasive cardiovascular imaging and future cath-lab decision support.

Text-driven medical image segmentation demands unified models based on multi-modal learning, capable of processing diverse imaging modalities while providing spatially reliable prediction estimates for clinical practice. Nevertheless, existing vision-language model solutions suffer from poor domain generalization and exhibit limited capability for uncertainty estimation to model the cumulative unreliability of cross-modal alignment across network layers. In this paper, we propose UniMedSeg, a unified vision-language framework for multi-modal medical image segmentation built upon frozen pre-trained vision-language backbones. Our framework incorporates modality-aware adaptation, cross-layer uncertainty-guided encoding, and a frequency-spatial dual-prompt decoder, wherein the progressive uncertainty propagation mechanism of the encoder suppresses misleading cross-modal features, whereas the decoder disentangles anatomical structure and boundary details in the frequency domain under dual textual prompts. Auxiliary losses further sharpen boundary representations and calibrate pixel-wise uncertainty maps. Extensive experiments demonstrate that UniMedSeg preserves competitive in-domain segmentation performance, improves domain generalization, and yields clinically useful spatial risk estimation.

Purpose: To develop and evaluate a recurrent neural network that operates directly on undersampled golden-angle radial k-space, enabling low-latency cardiac cine reconstruction for interventional cardiac MRI (iCMR). Methods: KOSTAL (K-space Optimized Spatial Temporal Architecture for Low latency) combines a per-coil k-space convolutional gated recurrent unit (convGRU), which learns a mask-conditioned view-sharing rule over the accumulated k-space, with an image-space recurrent module. It was trained and tested on the open-access multi-coil OCMR dataset using 10 simulated golden-angle spokes per frame against a 1000-spoke reference. Ablations varied recurrent module architecture and placement; temporal behavior was assessed using a imaging-plane transition test. Generalization was evaluated without retraining in 23 pediatric patients prospectively scanned during free-breathing using a radial golden-angle bSSFP sequence with binned XD-GRASP and clinical breath-held cine serving as references. Results: At 10 spokes per frame, KOSTAL reached a median SSIM of 0.907 (gridding: 0.622) in under 34 ms of processing time per frame. Image quality (IQ) matched gridded results formed from 7-11 times more spokes. Recurrence in k-space rather than image space accounted for most of this gain (0.891 versus [≤]0.650). IQ recovered from an abrupt plane transition in 75 ms, versus 303 ms for 200-spoke gridding. KOSTAL had the best agreement with the binned reference and the highest long-axis CNR. KOSTAL accurately tracked left-ventricular blood-pool volume (end-diastolic volume bias -0.7 mL) with a shorter lag (71 ms) than iterative reconstruction (239-287 ms). Conclusion: Learned, mask-conditioned recurrence in k-space yields high image quality at a latency compatible with interventional guidance.

With the rapid development of modern biotechnology, DNA vectors have become fundamental tools for inserting, transferring, and expressing specific gene sequences in various fields such as gene cloning, gene expression, gene editing, and gene therapy. However, when dealing with complex structured DNA sequences, traditional vector construction methods face challenges with low connection efficiency. This study proposes a new method for constructing recombinant vectors by employing a strategy of high-temperature treatment followed immediately by placement on ice, effectively reducing the complexity of DNA structures and enhancing the efficiency of PCR product-vector connection, thereby improving the construction efficiency of recombinant vectors. This paper describes the technical details of the method, experimental validation, and applications in gene cloning, gene recombination editing, and the preparation of gene therapy drugs, providing a new efficient tool for molecular biology experiments.

Monkeypox, caused by MPXV, has become a global public health concern; however, the T-cell immunity against MPXV following smallpox vaccination remains unclear. In this study, we screened the immunogenicity of 21 predicted MPXV epitopes and compared immune responses between young and elderly groups. Two immunogenic CD8 T-cell epitopes, M15 and M17, were identified, both of which effectively stimulated specific immune responses, with significantly stronger responses observed in the young group. Notably, this study provides the first evidence that smallpox vaccination induces CD8 T-cell immunity that cross-recognizes the MPXV epitopes M15 and M17 and remains detectable for more than four decades after vaccination, as demonstrated in vaccinated individuals aged 53-75. These findings provide experimental evidence for understanding the cross-protective mechanisms of the smallpox vaccine and informing monkeypox vaccine development.

FluPep is a 16 amino-acid peptide with potent anti-viral activity against influenza A virus, including the H1N1 subtype. To understand how FluPep exerts its effects, gold nanoparticles were functionalised with FluPep, since gold nanoparticles provide a very sensitive probe that can be detected optically. Our results demonstrate that nanoparticle-FluPep conjugates did not bind to influenza virus particles in vitro. However, the nanoparticle-FluPep conjugates bound to polyanions on the cell surface, and after washing cells they remained cell-associated and were able to inhibit flu virus infectivity. Nanoparticle-FluPep conjugates bound to heparin affinity columns and their binding to cells could be reduced by pre-treatment of the cells with heparinase and neuraminidase, but not with chondroitinase ABC. These data indicated that FluPep bound to heparan sulfate and sialic acid-containing glycans in the pericellular matrix and in doing so prevented viral infectivity. The importance of the interaction of FluPep with heparan sulfate (HS) was examined by extending FluPep at its C-terminus with amino acid sequences responsible for the interactions of growth factors with heparin. The resultant two new engineered peptides, termed superFluPep1 and superFluPep2, were more potent than parental FluPep. The results demonstrate that FluPep likely exerts its antiviral activity by interfering with virus-cell interactions and that enhancing the binding of such peptides to heparan sulfate is an effective strategy to increase their anti viral potency.

Development of gene therapies for USH2A-associated diseases has been limited by the large size of the USH2A coding sequence, which exceeds the packaging capacity of adeno-associated viral vectors. Here, we evaluated high-capacity adenoviral vectors (HC-AdVs) as vehicles for a mutation-independent USH2A gene supplementation strategy in USH2A human retinal organoids. First, we generated human induced pluripotent stem cell (hiPSC) lines carrying a nonsense mutation in USH2A exon-61 (61KO). Similarly to our previously established exon-13 knockout model (13KO), 61KO hiPSC-derived retinal organoids showed reduced levels of usherin and its interacting proteins, adhesion G protein-coupled receptor V1 (ADGRV1), and whirlin, at the photoreceptor connecting cilium. Delivery of a HC-AdV5 encoding full-length USH2A under the control of the photoreceptor-specific human rhodopsin kinase (hGRK1) promoter restored usherin expression and promoted re-localization of ADGRV1 and whirlin to their physiological sites at the connecting cilium in 13KO and 61KO photoreceptors. Unexpectedly, usherin expression was also observed within mutant Muller glial cells. These findings provide proof-of-concept that HC-AdV5-mediated delivery of full-length USH2A can successfully promote the rescue of usherin and relocate its interacting partners in USH2A-defective retinal organoids, demonstrating its potential as a variant-independent therapeutic approach for inherited retinal disorders caused by USH2A mutations.

IntroductionWhile placental perfusion and pathology jointly affect pregnancy outcomes, cotyledon-specific perfusion across gestation and its correlation with local injury is not yet well understood. Ferumoxytol dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) offers a promising way to noninvasively identify cotyledons across gestation and quantify longitudinal cotyledon-specific perfusion changes. Additionally, intraplacental injection of bioactive fibrin sealant allows us to model thrombotic placental injury and further assess cotyledon-level relationships between perfusion and significant injury. MethodsPregnant rhesus macaques (N=13) received intrauterine saline or fibrin sealant injections at gestational day (GD) [~]101 and underwent ferumoxytol DCE-MRI at GDs [~]100, 115, and 145. Placental perfusion domains derived from contrast arrival time were segmented at each imaging time point and matched to cotyledons identified following tissue collection by cesarean section, with cotyledon perfusion quantified longitudinally and correlated with cotyledon-specific quantitative histopathology. ResultsAll pregnancies were successfully carried to term. Fibrin sealant injections induced significantly higher levels of placental pathology compared to saline controls. MRI-derived perfusion domains were largely consistent across gestation and showed predominantly one-to-one correspondence with term cotyledons, with successful perfusion-pathology pairing achieved in 153 cotyledons. Longitudinal cotyledon perfusion changes showed significant positive correlations with villous agglutination injuries. ConclusionsFeasibility of noninvasively tracking placental cotyledon perfusion using ferumoxytol DCE-MRI was demonstrated, and the efficacy of the rhesus macaque thrombotic injury model was confirmed. The positive perfusion-pathology correlations suggested intrinsic placental regulatory mechanisms and functional plasticity. This new framework is promising for future translational studies and validation of ex vivo cotyledon perfusion models. HighlightsO_LILongitudinal tracking of placental perfusion domains with ferumoxytol MRI C_LIO_LISuccessful matching of cotyledons and MRI-derived perfusion domains C_LIO_LIConfirmed thrombotic injury-model induced cotyledon pathology C_LIO_LIMaternal perfusion compensation in presence of villous pathology C_LI

Connected speech is essential for everyday communication, but clinical constraints and patient fatigue limit detailed evaluation in acute stroke (<1-week post-stroke). Bedside assessments may sample discourse but rarely quantify language impairment (LI) in connected speech, leaving patient communication poorly characterized. We analyzed brief story retellings from 86 patients with left-hemisphere stroke (~4 days post-stroke; 63 classified with LI using composite clinical and naming criteria). From transcripts generated with automatic speech recognition, we derived discrete linguistic features and embeddings with Large Language Models (LLMs). An ensemble of embedding-based classifiers distinguished patients with and without LI with 90% balanced accuracy (79% sensitivity, 100% specificity), outperforming independent embedding and discrete-linguistic-based classifiers, showing distinct LLMs contributed complementary information. Adding the discrete-linguistic-based classifier to the ensemble did not improve balanced accuracy but modestly increased sensitivity at the expense of specificity. We provide proof of concept for a fast, largely automated discourse screener of acute LI.

Kyasanur Forest Disease Virus (KFDV) NS5 methyltransferase (MTase) protein is the essential enzyme that is involved in the cap methylation of viral RNA, viral replication, and immune evasion, and therefore it is an important protein of interest for antiviral research and drug design. In the present work, we successfully resolved the three-dimensional crystal structures of KFDV NS5 MTase co-crystallised with SAH and GTP at resolutions of 2.2 [A] and 2.6 [A], respectively. In previous studies, HC (Herbacetin) and CAPE (Caffeic acid phenethyl ester) have shown inhibitory activity against SAM-dependent viral MTase. To evaluate the inhibitory potential of HC and CAPE against KFDV NS5 MTase, we have performed isothermal titration calorimetry (ITC) and tryptophan fluorescence spectroscopy (TFS) to validate protein interaction with target compounds. MTase inhibition assay was performed using capillary electrophoresis (CE) assays. Additionally, fluorescence polarisation (FP) confirmed RNA binding inhibition by CAPE and HC. Together, these experiments suggest that HC and CAPE are promising inhibitors against KFDV NS5 MTase and could potentially act as lead compounds to design broad-spectrum anti-Orthoflavivirus drugs.

Quantitative assessment of head growth supports early detection of neurological disease, yet current clinical practice relies on manual measurements that are variable, anatomically limited, and difficult to scale. Here we introduce a fully automated imaging-native framework (CranioTrace) that transforms routine neuroimaging into standardized, population-referenced markers of cranial development. By integrating topology-constrained segmentation, contour regularization, and geometry-based quality control for automated slice selection, CranioTrace robustly extracts head circumference and cranial morphology from heterogeneous MRI and CT data without manual intervention. The framework enabled construction of sex-specific population growth models across pediatric development in a dataset comprising 9,685 scans spanning birth to early adulthood across 25 cohorts. Imaging-derived head circumference shows strong agreement with tape measurements across independent datasets, high reproducibility (intraclass correlation up to 0.99), and consistent performance across modalities. Beyond conventional head circumference, CranioTrace quantifies cranial shape and asymmetry, capturing developmental dynamics that are not accessible through conventional measurements. Population modeling reveals rapid early-life expansion followed by nonlinear deceleration and stable sex differences. Application to neurogenetic cohorts identifies disease-consistent shifts in growth trajectories: head circumference was higher in neurofibromatosis type 1 and 16p11.2 deletion and lower in 22q11.2 deletion and 16p11.2 duplication. By converting clinical imaging archives into scalable cranial phenotypes linked to probabilistic reference models, CranioTrace provides a foundation for imaging-based growth charting, integrated skull-brain phenotyping, and precise assessment of neurodevelopment.

Cholesterol is a key component of eukaryotic cell membranes, promoting membrane stability and modulating the function of many membrane proteins, including ion channels. In our previous work using purified human voltage-gated proton channel proteins, we showed that cholesterol inhibits the hHv1 channel by altering the conformational dynamics of its S4 segment, the key element that senses membrane voltage to control proton permeation. In the present work, we examined the effects of cholesterol analogs and potential sites in the hHv1 channel mediating cholesterol inhibition using site-directed mutagenesis and docking simulations. Our results showed that desmosterol, the immediate precursor of cholesterol, markedly attenuates cholesterol inhibition. Using single-molecule Fluorescence Resonance Energy Transfer (smFRET), we showed that desmosterol attenuates cholesterol inhibition by promoting the intermediate and open state conformations of the S4 segment. Moreover, we identified multiple residues in the hHv1 channel that are critical for cholesterol inhibition, including Y141A in the S2 segment, which reduces cholesterol inhibition by nearly 3-fold. Our smFRET results showed that the Y141A mutation promotes the intermediate conformation in the S4 segment, which underlies the attenuation of cholesterol inhibition. Consistently, docking simulations also revealed multiple residues spanning the transmembrane domain, rather than clustered within a single localized pocket. Our work identified the key molecular determinant in the hHv1 channel that mediates cholesterol inhibition and also provided a mechanism linking the conversion between demosterol and cholesterol by DHCR24 to pH homeostasis in many cells, such as phagocytes, cardiomyocytes, neurons and microglial cells.

Echovirus 18 (E18) is a predominant pathogen causing aseptic meningitis in children, and post-E18 infection frequently triggers myasthenia gravis-like autoimmune neurological damage. This pathological process relies on neonatal Fc receptor (FcRn)-mediated IgG transcytosis across mucosal barriers, and FcRn also acts as an essential functional receptor required for E18 attachment and uncoating during host cell invasion. At present, no E18-specific prophylactic vaccine has been clinically approved, and anti-FcRn monoclonal antibodies are the available therapeutics to alleviate autoantibody-mediated tissue injury. We constructed an integrated automated phylogenetic pipeline named evolution_conservation, which enables rapid tracing of the evolutionary position and genetic relatedness of clinical isolates to identify closely related strains from previous outbreaks. Serving as an in silico alternative to animal experiments, this pipeline supports reference-guided vaccine design and longitudinal comparative assessment of vaccine safety and efficacy, facilitates identification of patient populations presenting rare post-viral sequelae, and accelerates clinical trial progression. In this study, we inserted the pre-screened linear epitope E18-3 into a truncated hepatitis B core (HBc) scaffold to generate chimeric virus-like particles (VLPs). A non-invasive floss-based gingival mucosal immunization mouse model was established, with subcutaneous Freunds adjuvant immunization set as the control group. ELISA results confirmed that gingival mucosal delivery of particulate HBc-E18-3 VLPs alone could induce sustained high levels of antigen-specific intestinal IgG in vivo. Drawing on research paradigms of therapeutic neoantigen vaccines for tumor recurrence prevention, the evolution_conservation bioinformatic pipeline and mucosal VLP platform described herein establish an innovative framework for developing antigen-competitive prophylactic and therapeutic vaccines targeting FcRn for myasthenia gravis and autoimmune encephalitis.

Influenza virus infects the respiratory mucosa, highlighting the importance of mucosal immunity for early protection and transmission control. Here, we evaluated whether intranasal (IN) vaccination with recombinant trimeric hemagglutinin protein from A/Michigan/45/2015 (triHA) formulated with a combined mucosal adjuvant, nanoemulsion plus IVT, an RNA-based RIG-I agonist (NE/IVT), could protect guinea pigs against heterologous A/Netherlands/602/2009 challenge and reduce viral transmission. To compare mucosal and parenteral immunization, IN triHA/NE/IVT was benchmarked against IN triHA alone, IM triHA/AddaVax (IM triHA/Advx), and standard IM quadrivalent inactivated influenza vaccine (QIV). We also tested whether IN triHA/NE/IVT could boost IM QIV- primed immunity and included animals previously infected with A/Michigan/45/2015 to model pre-existing infection- induced immunity. Transmission was assessed by co-housing naive sentinels with vaccinated, challenged donors. IN triHA/NE/IVT induced systemic humoral responses comparable to IM triHA/Advx while generating superior nasal mucosal IgA responses. Unexpectedly, IM triHA/AddaVax also induced detectable, albeit lower, mucosal IgG and IgA, contrasting with prior mouse data and highlighting species-specific differences. IN triHA/NE/IVT boosting after IM QIV enhanced serum IgG and mucosal IgA compared with QIV prime-boost alone and increased cross-neutralizing activity against antigenically distinct A/Victoria/4897/2022. Both IN triHA/NE/IVT and IN Michigan/15 prior- infection prevented detectable viral shedding after challenge, and naive sentinels co-housed with IN triHA/NE/IVT- vaccinated donors remained seronegative. Together, these findings support NE/IVT as a potential mucosal platform capable of inducing robust systemic and mucosal immunity and boosting IM vaccine-primed responses.

Most recessive inherited retinal dystrophies are, in principle, amenable to gene augmentation, yet only one such therapy has received regulatory approval. We sought to identify tractable autosomal recessive retinitis pigmentosa (RP) genes that could be treated using existing non-proprietary adeno-associated virus (AAV) components. More than 100 RP-associated genes were prioritized according to cellular expression, coding-sequence size, and the availability of cell type-specific promoters, yielding 19 candidate genes expressed predominantly in rods and/or retinal pigment epithelium (RPE). In rhesus monkey eyes, the human RHO and BEST1 promoters drove rod- and RPE-specific expression, respectively, whereas the GFAP and RLBP1 promoters were limited by absent Muller glial expression or dose-associated RPE toxicity. We selected PDE6B as a proof-of-concept gene and evaluated AAV8-RHO-PDE6B after neonatal subretinal delivery. The vector effectively preserved outer nuclear layer structure, electroretinography responses, and visually guided behaviors in rd1 and rd10 mice for at least 6 months. Dose-ranging studies identified retinal abnormalities at the higher doses, whereas the lower doses were comparatively well tolerated in eyes of mice and rhesus monkeys. Together, these findings define a preclinical framework for developing gene-augmentation vectors for a subset of early-stage autosomal recessive RP patients.

Deep learning methods, such as AlphaFold and RosettaFold, achieve high accuracy in protein structure prediction. However, predicting the structure of large protein complexes remains challenging due to their large size and intricate multi-chain interactions. Docking-based methods can handle large proteins, but are limited by the huge combinatorial binding space of multichains. Assembly-based approaches offer an alternative, but their accuracy critically relies on the precision of predicted subcomponents. Addressing the challenges, we propose HDOCK-Multimer (HDM), a structure prediction framework of large protein complexes by integrating ab initio docking and combinatorial assembly. HDM can efficiently reduce reliance on subcom-ponent accuracy through docking process, while leveraging the pairwise interactions of subcom-ponents through assembly strategy. HDM is extensively validated on three benchmarks of 35 large heteromeric complexes, 172 large protein complexes, and 7 CASP15 targets, and compared with state-of-the-art methods including MoLPC, CombFold, AlphaFold-Multimer (AFM), and AlphaFold3 (AF3). It is shown that HDOCK-Multimer substantially outperforms the other methods. In addition, HDM also shows ability to predict the stoichiometry and model the complex without stoichiometry input. It is anticipated that HDM will serve as a powerful tool for studying large protein complexes or molecular machines. The HDM package is freely available at https://github.com/huang-laboratory/HDOCK-Multimer/.

The subgenual anterior cingulate cortex (sgACC) is a key node in treatment-resistant depression (TRD), but precise non-invasive neuromodulation of this target is challenging. Preclinical studies of non-ablative stereotactic radiosurgery (SRS) have shown neuromodulatory ("radiomodulation") effects. In this single-center, double-masked, randomized, dose-seeking pilot trial, nine adults with TRD were randomly assigned to bilateral sgACC radiomodulation at a dose of either 15, 20, or 25 Gy per hemispheric target. Primary endpoints were safety and feasibility; the efficacy endpoint was week-4 change in the Montgomery-Asberg Depression Rating Scale (MADRS). Both primary endpoints were met: the only treatment-related adverse event was transient grade 1 dizziness, with no structural MRI abnormality through week 12. Mean MADRS fell from 33.0 to 17.0 (48.5% reduction); 67% responded and 44% remitted, with benefit sustained to week 12. Resting-state fMRI revealed regional connectivity changes correlating with clinical improvement, with tractography showing streamline counts differing by response status. These first-in-human findings support a larger randomized controlled trial of sgACC radiomodulation for TRD. ClinicalTrial.gov registration: NCT07274917.