Citations

Abstract Genome-resolved virome analysis remains inaccessible for many samples, including those with clinical relevance, because viral nucleic acid recovered after enrichment is often too scarce to support de novo genome assembly. As a result, many analyses are limited to sparse read-level detection, which cannot recover divergent viruses, resolve strains, or interpret gene-level variation. Here, we developed Low Biomass Viral Sequencing (LBV-Seq), a workflow that couples low-input viral sample handling with modified primary template-directed amplification and short- or long-read sequencing to enable de novo reconstruction of DNA and RNA viral genomes from sub-femtogram to nanogram inputs. LBV-Seq reproducibly captures the same relative community composition, amplifies diverse viruses, and achieved broad genome coverage across nearly all targets regardless of viral genome structure, Baltimore class, abundance, or input mass. Short-read assemblies recovered near-complete genomes from femtogram-scale inputs. Long-read sequencing provided orthogonal support for genome structure, with PacBio HiFi reads spanning large portions of viral genomes and, in some cases, complete small viral genomes. Applied to virus-enriched human duodenal biopsy eluates, LBV-Seq provided proof-of-feasibility for recovering both bacteriophage and eukaryotic viral genomes from low-input biopsy-derived material. In the eluates tested, LBV-Seq recovered co-occurring Alphatorquevirus and Betatorquevirus genomes estimated to be present at roughly 10 copies/µL and at viral masses below 0.1 fg/µL. LBV-Seq enables genome-resolved virome analysis in samples previously limited to detection-based viromics, supporting viral discovery and strain-resolved analyses in settings where viral mass is low, including viruses enriched from human tissue biopsies.

Abstract Alzheimer’s Disease (AD) 1 is a progressive neurodegenerative disorder with limited therapeutic options, driving interest in drug repurposing to accelerate treatment discovery. Drug repurposing has emerged as a promising strategy to accelerate therapeutic discovery by repositioning existing drugs for new clinical indications. Recent computational repurposing approaches, including knowledge graph reasoning, transcriptomic signature analysis, and integrative literature mining, have demonstrated strong predictive capabilities 2 . However, these methods often yield divergent drug rankings, which makes it difficult to decide which candidates to advance for experimental follow-up and results in substantial gaps between computational predictions and feasible in vivo validation 2 .To bridge this computational-experimental gap, we proposed an advanced prioritization framework leveraging large language models (LLMs). Our method systematically evaluated three state-of-the-art (SOTA) and representative computational methods (TxGNN 3 , Composition-based Graph Convolutional Network (CompGCN) 4 , and a regularized logistic regression (RLR) 5 , to analyze both their predictive performance and pharmaceutical class distributions. By integrating the strengths and divergences of these models, we generated a unified, streamlined list of 90 candidate drugs for further prioritization. We then utilized an LLM-based agent to perform evidence synthesis from biomedical literature abstracts for each candidate. This process mimics expert manual curation but significantly reduces human effort and time by efficiently distilling vast textual data into actionable insights. Applying consistent and transparent selection criteria, we obtained a refined and prioritized list of drug candidates suitable for subsequent in vivo experimental validation. The robustness and clinical relevance of our framework were validated using real-world data from Alzheimer’s patient cohorts, clinical trial registries, and expert pharmacological reviews. This comprehensive validation confirmed that our LLM-driven approach enhances efficiency, consistency, scalability, and generalizability. By integrating computational predictions with scalable evidence synthesis and multifaceted validation, our framework facilitated rapid and informed prioritization of repurposed drugs. Our framework can potentially accelerate the translational pathway toward viable AD therapeutics. Moreover, the versatility of our framework can also be applied to drug repurposing efforts for other diseases beyond AD.

e19025 Background: Recent studies suggest that patients with CLL/SLL (herein, CLL) relapsed/refractory (R/R) to both cBTKi and BCL2i (double R/R [DR]) may have poorer outcomes than the broader group of double-exposed (DE) patients. This study evaluated real-world response (rwR) and time-to-event (TTE) outcomes of a CLL DE cohort and DR subcohort in the US. Methods: This retrospective study used the nationwide Flatiron Health electronic health record-derived deidentified database to include patients with CLL treated with sequential or combination cBTKi and BCL2i in any line of therapy (LOT) on or after January 1, 2011, and a subsequent LOT (ie, index LOT) on or before November 30, 2022 (DE cohort). A subset with DR CLL (DR subcohort) was analyzed separately. Descriptive statistics were generated for demographic and clinical characteristics. Outcomes included clinician-confirmed rwR rate (rwRR) and median TTE outcomes (real-world duration of response [rwDOR], progression-free survival [rwPFS], time to next treatment [rwTTNT], and overall survival [rwOS]; Kaplan Meier, follow-up to May 31, 2023). Results: The DE cohort included 264 patients, 107 of whom were in the DR subcohort. Patients were mostly male (64%), White (75%), aged ≥65 years at first index LOT (72%) after prior cBTKi and BCL2i treatment, and treated in the community setting (71%). Approximately one-third of patients received a subsequent treatment after each index LOT. Most patients were re-treated with a cBTKi and/or BCL2i in the first index LOT (69% DE; 54% DR), and the percentage treated with chemotherapy increased with each subsequent index LOT. Confirmed rwRR was similar between DE and DR for all index LOTs (Table). More relevant are the median TTE outcomes, which were shorter for DR than DE, especially for the first index LOT, and shorter in subsequent index LOTs for both groups. Conclusions: The high percentage of patients re-treated with cBTKi and BCL2i therapies and increasing use of chemotherapy regimens with each subsequent index LOT indicate a lack of standard of care and options. Our results suggest a high unmet need in the DE CLL population, particularly in patients DR to cBTKi and BCL2i therapy, who have worse outcomes. Efficacious and durable therapy options are needed to address this unmet need. Double Exposed Index LOT Double R/R Index LOT 1st 2nd 3rd 1st 2nd 3rd N 264 90 35 107 41 12 rwRR, % (95% CI) 39(33-45) 42(32-53) 40(24-58) 36(27-45) 34(21-51) 42(16-71) Median TTE outcomes, months (95% CI) rwDOR 20.3 (13.5-32.6) 6.0(5.5-NR) 3.1(2.7-NR) 11.5 (7.9-23.5) 6.0(3.6-NR) 2.9(2.7-NR) rwPFS 8.8(7.5-10.6) 5.1(4.0-7.3) 4.3(2.9-6.3) 5.0 (4.0-7.3) 3.6(2.6-5.7) 4.7(2.7-NR) rwTTNT 13.8 (10.3-18.3) 7.9(5.7-11.3) 5.7(4.2-14.6) 8.0 (6.2-11.5) 5.9(4.4-11.4) 5.0(4.2-NR) rwOS 25.2 (22.1-37.1) 15.2(11.4-22.6) 6.6(5.7-NR) 14.7 (12.3-22.6) 11.4(5.9-NR) 6.6(5.0-NR)

AbstractAutoimmune pancreatitis (AIP) is identified as a severe chronic immune‐related disorder in pancreas, including two subtypes. In this study, pancreatic lesions in patients diagnosed as either type 1 AIP or type 2 AIP are examined, and these patients’ peripheral blood at single‐cell level. Furthermore, flow cytometry, immunofluorescence, and functional assays are performed to verify the identified cell subtypes. In type 1 AIP, there is a notable increase in the amount of B cells and plasma cells, and IgG4+ plasma cells are key pathogenic cells of AIP. The differentiation path of naïve‐stage B cells into IgG4+ produced plasma cells is observed, and an increased amount of T helper cells and T follicular helper (Tfh) cells. This study also reveals that HIF‐1α, an activated transcriptional factor, can directly bind to promoter site of NAMPT, promoting higher levels of visfatin production in HIF1A+ classical monocytes. Pancreatic stellate cells can be activated by extracellular visfatin and promote the development of fibrotic response in pancreatic lesions across both AIP subtypes. The current findings shed light on the exploration of dynamic alterations in peripheral blood cells and cell subgroups in pancreatic lesions of AIP, while elucidating a pathogenic cell subset and potential fibrosis mechanism of AIP.

Abstract Objectives Precise literature recommendation and summarization are crucial for biomedical professionals. While the latest iteration of generative pretrained transformer (GPT) incorporates 2 distinct modes—real-time search and pretrained model utilization—it encounters challenges in dealing with these tasks. Specifically, the real-time search can pinpoint some relevant articles but occasionally provides fabricated papers, whereas the pretrained model excels in generating well-structured summaries but struggles to cite specific sources. In response, this study introduces RefAI, an innovative retrieval-augmented generative tool designed to synergize the strengths of large language models (LLMs) while overcoming their limitations. Materials and Methods RefAI utilized PubMed for systematic literature retrieval, employed a novel multivariable algorithm for article recommendation, and leveraged GPT-4 turbo for summarization. Ten queries under 2 prevalent topics (“cancer immunotherapy and target therapy” and “LLMs in medicine”) were chosen as use cases and 3 established counterparts (ChatGPT-4, ScholarAI, and Gemini) as our baselines. The evaluation was conducted by 10 domain experts through standard statistical analyses for performance comparison. Results The overall performance of RefAI surpassed that of the baselines across 5 evaluated dimensions—relevance and quality for literature recommendation, accuracy, comprehensiveness, and reference integration for summarization, with the majority exhibiting statistically significant improvements (P-values <.05). Discussion RefAI demonstrated substantial improvements in literature recommendation and summarization over existing tools, addressing issues like fabricated papers, metadata inaccuracies, restricted recommendations, and poor reference integration. Conclusion By augmenting LLM with external resources and a novel ranking algorithm, RefAI is uniquely capable of recommending high-quality literature and generating well-structured summaries, holding the potential to meet the critical needs of biomedical professionals in navigating and synthesizing vast amounts of scientific literature.