Hasil untuk "Internal medicine"

Jianmin Li, Ying Chang, Su-Kit Tang et al.

Background: Retrieval augmented generation (RAG) technology can empower large language models (LLMs) to generate more accurate, professional, and timely responses without fine tuning. However, due to the complex reasoning processes and substantial individual differences involved in traditional Chinese medicine (TCM) clinical diagnosis and treatment, traditional RAG methods often exhibit poor performance in this domain. Objective: To address the limitations of conventional RAG approaches in TCM applications, this study aims to develop an improved RAG framework tailored to the characteristics of TCM reasoning. Methods: We developed TCM-DiffRAG, an innovative RAG framework that integrates knowledge graphs (KG) with chains of thought (CoT). TCM-DiffRAG was evaluated on three distinctive TCM test datasets. Results: The experimental results demonstrated that TCM-DiffRAG achieved significant performance improvements over native LLMs. For example, the qwen-plus model achieved scores of 0.927, 0.361, and 0.038, which were significantly enhanced to 0.952, 0.788, and 0.356 with TCM-DiffRAG. The improvements were even more pronounced for non-Chinese LLMs. Additionally, TCM-DiffRAG outperformed directly supervised fine-tuned (SFT) LLMs and other benchmark RAG methods. Conclusions: TCM-DiffRAG shows that integrating structured TCM knowledge graphs with Chain of Thought based reasoning substantially improves performance in individualized diagnostic tasks. The joint use of universal and personalized knowledge graphs enables effective alignment between general knowledge and clinical reasoning. These results highlight the potential of reasoning-aware RAG frameworks for advancing LLM applications in traditional Chinese medicine.

Youngseung Jeon, Christopher Hwang, Ziwen Li et al.

While drug discovery is vital for human health, the process remains inefficient. Medicinal chemists must navigate a vast protein space to identify target proteins that meet three criteria: physical and functional interactions, therapeutic impact, and docking potential. Prior approaches have provided fragmented support for each criterion, limiting the generation of promising hypotheses for wet-lab experiments. We present HAPPIER, an AI-powered tool that supports hypothesis generation with integrated multi-criteria support for target identification. HAPPIER enables medicinal chemists to 1) efficiently explore and verify proteins in a single integrated graph component showing multi-criteria satisfaction and 2) validate AI suggestions with domain knowledge. These capabilities facilitate iterative cycles of divergent and convergent thinking, essential for hypothesis generation. We evaluated HAPPIER with ten medicinal chemists, finding that it increased the number of high-confidence hypotheses and support for the iterative cycle, and further demonstrated the relationship between engaging in such cycles and confidence in outputs.

Michael S. Yao, Osbert Bastani, Alma Andersson et al.

The goal of personalized medicine is to discover a treatment regimen that optimizes a patient's clinical outcome based on their personal genetic and environmental factors. However, candidate treatments cannot be arbitrarily administered to the patient to assess their efficacy; we often instead have access to an in silico surrogate model that approximates the true fitness of a proposed treatment. Unfortunately, such surrogate models have been shown to fail to generalize to previously unseen patient-treatment combinations. We hypothesize that domain-specific prior knowledge - such as medical textbooks and biomedical knowledge graphs - can provide a meaningful alternative signal of the fitness of proposed treatments. To this end, we introduce LLM-based Entropy-guided Optimization with kNowledgeable priors (LEON), a mathematically principled approach to leverage large language models (LLMs) as black-box optimizers without any task-specific fine-tuning, taking advantage of their ability to contextualize unstructured domain knowledge to propose personalized treatment plans in natural language. In practice, we implement LEON via 'optimization by prompting,' which uses LLMs as stochastic engines for proposing treatment designs. Experiments on real-world optimization tasks show LEON outperforms both traditional and LLM-based methods in proposing individualized treatments for patients.

Daniel Mas Montserrat, Ray Verma, Míriam Barrabés et al.

Large-scale genomic workflows used in precision medicine can process datasets spanning tens to hundreds of gigabytes per sample, leading to high memory spikes, intensive disk I/O, and task failures due to out-of-memory errors. Simple static resource allocation methods struggle to handle the variability in per-chromosome RAM demands, resulting in poor resource utilization and long runtimes. In this work, we propose multiple mechanisms for adaptive, RAM-efficient parallelization of chromosome-level bioinformatics workflows. First, we develop a symbolic regression model that estimates per-chromosome memory consumption for a given task and introduces an interpolating bias to conservatively minimize over-allocation. Second, we present a dynamic scheduler that adaptively predicts RAM usage with a polynomial regression model, treating task packing as a Knapsack problem to optimally batch jobs based on predicted memory requirements. Additionally, we present a static scheduler that optimizes chromosome processing order to minimize peak memory while preserving throughput. Our proposed methods, evaluated on simulations and real-world genomic pipelines, provide new mechanisms to reduce memory overruns and balance load across threads. We thereby achieve faster end-to-end execution, showcasing the potential to optimize large-scale genomic workflows.

Safwaan H. Khan, Safwaan H. Khan, Yeonjoo Choi et al.

Chimeric antigen receptor (CAR) T cell therapy has revolutionized the treatment of hematologic malignancies, achieving remarkable clinical success with FDA-approved therapies targeting CD19 and BCMA. However, the extension of these successes to solid tumors remains limited due to several intrinsic challenges, including antigen heterogeneity and immunosuppressive tumor microenvironments. In this review, we provide a comprehensive overview of recent advances in CAR T cell therapy aimed at overcoming these obstacles. We discuss the importance of antigen identification by emphasizing the identification of tumor-specific and tumor-associated antigens and the development of CAR T therapies targeting these antigens. Furthermore, we highlight key structural innovations, including cytokine-armored CARs, protease-regulated CARs, and CARs engineered with chemokine receptors, to enhance tumor infiltration and activity within the immunosuppressive microenvironment. Additionally, novel manufacturing approaches, such as the Sleeping Beauty transposon system, mRNA-based CAR transfection, and in vivo CAR T cell production, are discussed as scalable solution to improve the accessibility of CAR T cell therapies. Finally, we address critical therapeutic limitations, including cytokine release syndrome (CRS), immune effector cell-associated neurotoxicity syndrome (ICANS), and suboptimal persistence of CAR T cells. An examination of emerging strategies for countering these limitations reveals that CRISPR-Cas9-mediated genetic modifications and combination therapies utilizing checkpoint inhibitors can improve CAR T cell functionality and durability. By integrating insights from preclinical models, clinical trials, and innovative engineering approaches, this review addresses advances in CAR T cell therapies and their performance in solid tumors.

Rebecca Black, BApSc, Speech Pathologist(SP), Duy Duong Nguyen, MD PhD, Anna Miles, PhD et al.

Background: Oropharyngeal dysphagia and laryngeal dysfunction are complications of lung and heart transplantation. However, there is a lack of understanding around pre-operative function and an absence of standardized assessment protocols. We aimed to trial a pre- and post-operative protocol for assessing voice and swallowing function. Method: A prospective, longitudinal study of 14 adults undergoing investigation for lung or heart transplantation was conducted at a tertiary referral hospital. Patients were assessed pre-surgery and up to 6 months afterwards. The protocol involved phonation tasks with auditory-perceptual and acoustic analysis, videolaryngostroboscopy, a flexible endoscopic examination of swallowing and patient reported quality of life measures. Risk factors and clinical outcomes were extracted from patient records. Results: Patient self-reports of swallowing and voice difficulties were elevated pre-operatively. No evidence of swallowing difficulty was observed under endoscopic examination pre-transplant (Penetration-Aspiration Scale score <2; no accumulated secretions) and only one patient presented with incomplete glottic closure. Auditory perceptual ratings revealed voices were largely within the healthy range at baseline. One out of five patients presented with severe dysphonia post-operatively. Completion of evaluation measures prior to transplantation was 79% but post- operative rates were low due to feasibility challenges with follow up in this complex population. Conclusion: Novel evidence of self-reported pre-transplant voice and swallowing changes indicate value in baseline screening. Discrepancies between patient-report and instrumental assessment results highlight the need for multi-faceted evaluation. Large cohort studies are needed to determine the salient evaluation measures and time points for voice and swallowing assessment in this population.

Michael Vollenweider, Manuel Schürch, Chiara Rohrer et al.

Precision medicine has the potential to tailor treatment decisions to individual patients using machine learning (ML) and artificial intelligence (AI), but it faces significant challenges due to complex biases in clinical observational data and the high-dimensional nature of biological data. This study models various types of treatment assignment biases using mutual information and investigates their impact on ML models for counterfactual prediction and biomarker identification. Unlike traditional counterfactual benchmarks that rely on fixed treatment policies, our work focuses on modeling different characteristics of the underlying observational treatment policy in distinct clinical settings. We validate our approach through experiments on toy datasets, semi-synthetic tumor cancer genome atlas (TCGA) data, and real-world biological outcomes from drug and CRISPR screens. By incorporating empirical biological mechanisms, we create a more realistic benchmark that reflects the complexities of real-world data. Our analysis reveals that different biases lead to varying model performances, with some biases, especially those unrelated to outcome mechanisms, having minimal effect on prediction accuracy. This highlights the crucial need to account for specific biases in clinical observational data in counterfactual ML model development, ultimately enhancing the personalization of treatment decisions in precision medicine.

Deni Salja

Pseudocolimits are formal gluing constructions that combine objects in a category indexed by a pseudofunctor. When the objects are categories and the domain of the pseudofunctor is small and filtered it has been known since Exppose 6 in SGA4 that the pseudocolimit can be computed by taking the Grothendieck construction of the pseudofunctor and inverting the class of cartesian arrows with respect to the canonical fibration. This paper is a reformatted version of a MSc thesis submitted and defended at Dalhousie University in August 2022. The first part presents a set of conditions for defining an internal category of elements of a diagram of internal categories and proves it is the oplax colimit. The second part presents a set of conditions on an ambient category and an internal category with an object of weak-equivalences that allows an internal description of the axioms for a category of (right) fractions and a definition of the internal category of (right) fractions when all the conditions and axioms are satisfied. These are combined to present a suitable context for computing the pseudocolimit of a small filtered diagram of internal categories.

Hannes Jakob

We construct models in which there are stationarily many structures that exhibit different variants of internal approachability at different levels. This answers a question of Foreman. We also show that the approachability property at $μ$ is consistent with having a distinction of variants of internal approachability for stationarily many $N\in[H(μ^+)]^μ$. This is obtained using a new version of Mitchell Forcing.

Quan Wang, Yanmin Cao, Xinyu Liu et al.

Uppala Radhakrishna, Uppala Ratnamala, Devendrasinh D. Jhala et al.

BackgroundHidradenitis suppurativa (HS) is a chronic inflammatory skin condition that affects hair follicles in areas with apocrine sweat glands, such as the underarms, groin, and buttocks. The pathogenesis of HS is not fully understood, but considering the key role played by the biological clock in the control of immune/inflammatory processes the derangement of circadian and ultradian pathways could be hypothesized.MethodsWe analyzed genome-wide DNA methylation patterns in peripheral blood from 24 HS cases and 24 controls using the Infinium HumanMethylation450 BeadChip array (Illumina), followed by bioinformatics and statistical analyses.ResultsWe found that several circadian and ultradian genes were differentially methylated in HS patients, predominantly exhibiting hypomethylation. These genes were enriched in pathways such as MAPK and WNT cascades, acute phase response, cytokine release, inflammation, innate immune response, xenobiotic detoxification, and oxidative stress response.ConclusionAltered DNA methylation patterns of genes related to circadian and ultradian pathways could contribute to immune system derangement and inflammatory processes chronicization in addition to other comorbidities hallmarking HS onset and progression, at the same time representing possible druggable targets.

Joshua Wrigley

We study morphisms of internal locales of Grothendieck toposes externally: treating internal locales and their morphisms as sheaves and natural transformations. We characterise those morphisms of internal locales that induce surjective geometric morphisms and geometric embeddings, demonstrating that both can be computed `pointwise'. We also show that the co-frame operations on the co-frame of internal sublocales can also be computed `pointwise' too.

Hellmuth Weich, Philip Herbst, Francis Smit et al.

Rheumatic heart disease [RHD] is the most prevalent cause of valvular heart disease in the world, outstripping degenerative aortic stenosis numbers fourfold. Despite this, global resources are firmly aimed at improving the management of degenerative disease. Reasons remain complex and include lack of resources, expertise, and overall access to valve interventions in developing nations, where RHD is most prevalent. Is it time to consider less invasive alternatives to conventional valve surgery? Several anatomical and pathological differences exist between degenerative and rheumatic valves, including percutaneous valve landing zones. These are poorly documented and may require dedicated solutions when considering percutaneous intervention. Percutaneous balloon mitral valvuloplasty (PBMV) is the treatment of choice for severe mitral stenosis (MS) but is reserved for patients with suitable valve anatomy without significant mitral regurgitation (MR), the commonest lesion in RHD. Valvuloplasty also rarely offers a durable solution for patients with rheumatic aortic stenosis (AS) or aortic regurgitation (AR). MR and AR pose unique challenges to successful transcatheter valve implantation as landing zone calcification, so central in docking transcatheter aortic valves in degenerative AS, is often lacking. Surgery in young RHD patients requires mechanical prostheses for durability but morbidity and mortality from both thrombotic complications and bleeding on Warfarin remains excessively high. Also, redo surgery rates are high for progression of aortic valve disease in patients with prior mitral valve replacement (MVR). Transcatheter treatments may offer a solution to anticoagulation problems and address reoperation in patients with prior MVR or failing ventricles, but would have to be tailored to the rheumatic environment. The high prevalence of MR and AR, lack of calcification and other unique anatomical challenges remain. Improvements in tissue durability, the development of novel synthetic valve leaflet materials, dedicated delivery systems and docking stations or anchoring systems to securely land the transcatheter devices, would all require attention. We review the epidemiology of RHD and discuss anatomical differences between rheumatic valves and other pathologies with a view to transcatheter solutions. The shortcomings of current RHD management, including current transcatheter treatments, will be discussed and finally we look at future developments in the field.

Hewei Peng, Junchao Zhang, Xianhua Huang et al.

Abstract Background Nonalcoholic fatty liver disease (NAFLD), a common liver disease worldwide, can be reversed early in life with lifestyle and medical interventions. This study aimed to develop a noninvasive tool to screen NAFLD accurately. Methods Risk factors for NAFLD were identified using multivariate logistic regression analysis, and an online NAFLD screening nomogram was developed. The nomogram was compared with reported models (fatty liver index (FLI), atherogenic index of plasma (AIP), and hepatic steatosis index (HSI)). Nomogram performance was evaluated through internal and external validation (National Health and Nutrition Examination Survey (NHANES) database). Results The nomogram was developed based on six variables. The diagnostic performance of the present nomogram for NAFLD (area under the receiver operator characteristic curve (AUROC): 0.863, 0.864, and 0.833, respectively) was superior to that of the HSI (AUROC: 0.835, 0.833, and 0.810, respectively) and AIP (AUROC: 0.782, 0.773, and 0.728, respectively) in the training, validation, and NHANES sets. Decision curve analysis and clinical impact curve analysis presented good clinical utility. Conclusion This study establishes a new online dynamic nomogram with excellent diagnostic and clinical performance. It has the potential to be a noninvasive and convenient method for screening individuals at high risk for NAFLD.

Andrea Boccatonda, Lorenzo Andreetto, Damiano D’Ardes et al.

Non-alcoholic fatty liver disease (NAFLD) is defined as a chronic liver disease characterized by excessive fat accumulation in the liver without another obvious cause (no excessive alcohol consumption, hepatotoxic medications, toxins, viral infections, genetic hepatic diseases), therefore it is an exclusion diagnosis. The term NAFLD literally refers to non-alcohol related hepatopathy and does not adequately correlate with metabolic dysfunction and related cardiovascular risks. Therefore, researchers and scientific societies have moved towards changing the terminology. The novel nomenclature for a metabolic-associated fatty liver disease (MAFLD) has been proposed in 2020 by a group of experts to overcome the issues related to the old terminology. The diagnosis of MAFLD is based on the presence of hepatic steatosis and at least one between these three conditions: type 2 diabetes mellitus (T2DM), obesity or metabolic dysregulation. MAFLD has been shown to be an independent risk factor for cardiovascular diseases and atherosclerosis. It is better related to the main risk factors for atherosclerosis and cardiovascular diseases than NAFLD, such as dyslipidemia, T2DM and hypertension. The aim of this review is to highlight the reasons why the term NAFLD is moving to the term MAFLD, what are the conceptual basis of this choice and its clinical implications, particularly in the cardiovascular field.

Andrea Campagner, Lorenzo Famiglini, Anna Carobene et al.

In medical settings, Individual Variation (IV) refers to variation that is due not to population differences or errors, but rather to within-subject variation, that is the intrinsic and characteristic patterns of variation pertaining to a given instance or the measurement process. While taking into account IV has been deemed critical for proper analysis of medical data, this source of uncertainty and its impact on robustness have so far been neglected in Machine Learning (ML). To fill this gap, we look at how IV affects ML performance and generalization and how its impact can be mitigated. Specifically, we provide a methodological contribution to formalize the problem of IV in the statistical learning framework and, through an experiment based on one of the largest real-world laboratory medicine datasets for the problem of COVID-19 diagnosis, we show that: 1) common state-of-the-art ML models are severely impacted by the presence of IV in data; and 2) advanced learning strategies, based on data augmentation and data imprecisiation, and proper study designs can be effective at improving robustness to IV. Our findings demonstrate the critical relevance of correctly accounting for IV to enable safe deployment of ML in clinical settings.

James T. Brown, Anita Saigal, Nina Karia et al.

Background Ongoing exercise intolerance of unclear cause following COVID‐19 infection is well recognized but poorly understood. We investigated exercise capacity in patients previously hospitalized with COVID‐19 with and without self‐reported exercise intolerance using magnetic resonance–augmented cardiopulmonary exercise testing. Methods and Results Sixty subjects were enrolled in this single‐center prospective observational case‐control study, split into 3 equally sized groups: 2 groups of age‐, sex‐, and comorbidity‐matched previously hospitalized patients following COVID‐19 without clearly identifiable postviral complications and with either self‐reported reduced (COVIDreduced) or fully recovered (COVIDnormal) exercise capacity; a group of age‐ and sex‐matched healthy controls. The COVIDreducedgroup had the lowest peak workload (79W [Interquartile range (IQR), 65–100] versus controls 104W [IQR, 86–148]; P=0.01) and shortest exercise duration (13.3±2.8 minutes versus controls 16.6±3.5 minutes; P=0.008), with no differences in these parameters between COVIDnormal patients and controls. The COVIDreduced group had: (1) the lowest peak indexed oxygen uptake (14.9 mL/minper kg [IQR, 13.1–16.2]) versus controls (22.3 mL/min per kg [IQR, 16.9–27.6]; P=0.003) and COVIDnormal patients (19.1 mL/min per kg [IQR, 15.4–23.7]; P=0.04); (2) the lowest peak indexed cardiac output (4.7±1.2 L/min per m2) versus controls (6.0±1.2 L/min per m2; P=0.004) and COVIDnormal patients (5.7±1.5 L/min per m2; P=0.02), associated with lower indexed stroke volume (SVi:COVIDreduced 39±10 mL/min per m2 versus COVIDnormal 43±7 mL/min per m2 versus controls 48±10 mL/min per m2; P=0.02). There were no differences in peak tissue oxygen extraction or biventricular ejection fractions between groups. There were no associations between COVID‐19 illness severity and peak magnetic resonance–augmented cardiopulmonary exercise testing metrics. Peak indexed oxygen uptake, indexed cardiac output, and indexed stroke volume all correlated with duration from discharge to magnetic resonance–augmented cardiopulmonary exercise testing (P<0.05). Conclusions Magnetic resonance–augmented cardiopulmonary exercise testing suggests failure to augment stroke volume as a potential mechanism of exercise intolerance in previously hospitalized patients with COVID‐19. This is unrelated to disease severity and, reassuringly, improves with time from acute illness.

Tianxiang Sun, Yunhua Zhou, Xiangyang Liu et al.

As a simple technique to accelerate inference of large-scale pre-trained models, early exiting has gained much attention in the NLP community. It allows samples to exit early at internal classifiers without passing through the entire model. Most existing work usually trains the internal classifiers independently and employs an exiting strategy to decide whether or not to exit based on the confidence of the current internal classifier. However, none of these works takes full advantage of the fact that the internal classifiers are trained to solve the same task therefore can be used to construct an ensemble. In this paper, we show that a novel objective function for the training of the ensemble internal classifiers can be naturally induced from the perspective of ensemble learning and information theory. The proposed training objective consists of two terms: one for accuracy and the other for the diversity of the internal classifiers. In contrast, the objective used in prior work is exactly the accuracy term of our training objective therefore only optimizes the accuracy but not diversity. Further, we propose a simple voting-based strategy that considers predictions of all the past internal classifiers to infer the correct label and decide whether to exit. Experimental results on various NLP tasks show that our proposed objective function and voting-based strategy can achieve better accuracy-speed trade-offs.

Nobuya Inagaki

Chun-Yu Chen, Mohammed G Ghonime, Uksha Saini et al.

Background Oncolytic virotherapy (OV) is an immunotherapy that incorporates viral cancer cell lysis with engagement of the recruited immune response against cancer cells. Pediatric solid tumors are challenging targets because they contain both an inert immune environment and a quiet antigenic landscape, making them more resistant to conventional OV approaches. Further complicating this, herpes simplex virus suppresses host gene expression during virotherapy infection.Methods We therefore developed a multimodal oncolytic herpes simplex virus (oHSV) that expresses ephrin A2 (EphA2), a shared tumor-associated antigen (TAA) expressed by many tumors to improve immune-mediated antitumor activity. We verified the virus genotypically and phenotypically and then tested it in an oHSV-resistant orthotopic model (including immunophenotypic analysis), in flank and in T cell-deficient mouse models. We then assessed the antigen-expressing virus in an unrelated peripheral tumor model that also expresses the shared tumor antigen and evaluated functional T-cell response from the treated mice.Results Virus-based EphA2 expression induces a robust acquired antitumor immune responses in both an oHSV-resistant murine brain and peripheral tumor model. Our new multimodal oncolytic virus (1) improves survival in viroimmunotherapy resistant tumors, (2) alters both the infiltrating and peripheral T-cell populations capable of suppressing tumor growth on rechallenge, and (3) produces EphA2-specific CD8 effector-like populations.Conclusions Our results suggest that this flexible viral-based platform enables immune recognition of the shared TAA and improves the immune-therapeutic response, thus making it well suited for low-mutational load tumors.