
IEEE Access, Год журнала: 2024, Номер 12, С. 180652 - 180673
Опубликована: Янв. 1, 2024
Язык: Английский
IEEE Access, Год журнала: 2024, Номер 12, С. 180652 - 180673
Опубликована: Янв. 1, 2024
Язык: Английский
Bioinformatics, Год журнала: 2024, Номер 40(6)
Опубликована: Июнь 1, 2024
Abstract Motivation Answering and solving complex problems using a large language model (LLM) given certain domain such as biomedicine is challenging task that requires both factual consistency logic, LLMs often suffer from some major limitations, hallucinating false or irrelevant information, being influenced by noisy data. These issues can compromise the trustworthiness, accuracy, compliance of LLM-generated text insights. Results Knowledge Retrieval Augmented Generation ENgine (KRAGEN) new tool combines knowledge graphs, (RAG), advanced prompting techniques to solve with natural language. KRAGEN converts graphs into vector database uses RAG retrieve relevant facts it. techniques: namely graph-of-thoughts (GoT), dynamically break down problem smaller subproblems, proceeds each subproblem through framework, which limits hallucinations, finally, consolidates subproblems provides solution. KRAGEN’s graph visualization allows user interact evaluate quality solution’s GoT structure logic. Availability implementation deployed running its custom Docker containers. available open-source GitHub at: https://github.com/EpistasisLab/KRAGEN.
Язык: Английский
Процитировано
17Briefings in Bioinformatics, Год журнала: 2024, Номер 25(6)
Опубликована: Сен. 23, 2024
Feature selection in Knowledge Graphs (KGs) is increasingly utilized diverse domains, including biomedical research, Natural Language Processing (NLP), and personalized recommendation systems. This paper delves into the methodologies for feature (FS) within KGs, emphasizing their roles enhancing machine learning (ML) model efficacy, hypothesis generation, interpretability. Through this comprehensive review, we aim to catalyze further innovation FS paving way more insightful, efficient, interpretable analytical models across various domains. Our exploration reveals critical importance of scalability, accuracy, interpretability techniques, advocating integration domain knowledge refine process. We highlight burgeoning potential multi-objective optimization interdisciplinary collaboration advancing KG FS, underscoring transformative impact such on precision medicine, among other fields. The concludes by charting future directions, development scalable, dynamic algorithms explainable AI principles foster transparency trust KG-driven models.
Язык: Английский
Процитировано
4BioData Mining, Год журнала: 2025, Номер 18(1)
Опубликована: Фев. 18, 2025
Interdisciplinary, transdisciplinary, convergence, and No-Boundary Thinking (NBT) research are methodology technology-agnostic approaches to problem solving. The focus is on defining problems informed by access multiple knowledge sources expert perspectives across different domains, with the goal of accessing all available perspectives. While could be seen as a difficult attain objective, recent rise AI we might closer approaching this goal. We review several examples methodologies technologies that have been used put these strategies into action, but primary paper how advances in now enable quantum leap forward new problems. By leveraging capacity synthesize from tools can propose candidate definitions. uniquely able draw upon many more than any individual-or even very large team-could. Coupled human intelligence, better defined address complex scholarly or societal challenges.
Язык: Английский
Процитировано
0Current Issues in Molecular Biology, Год журнала: 2025, Номер 47(3), С. 200 - 200
Опубликована: Март 18, 2025
The rising prevalence of Alzheimer’s disease (AD), particularly among older adults, has driven increased research into its underlying mechanisms and risk factors. Aging, genetic susceptibility, cardiovascular health are recognized contributors to AD, but how the age onset affects progression remains underexplored. This study investigates role early- versus late-onset (EOAD LOAD, respectively) in shaping trajectory cognitive decline. Leveraging data from Religious Orders Study Memory Aging Project (ROSMAP), two cohorts were established: individuals with early-onset AD those AD. Comprehensive analyses, including differential gene expression profiling, pathway enrichment, co-expression network construction, conducted identify distinct molecular signatures associated each cohort. Network modularity learning algorithms used discern inner structure networks their related functional features. Computed descriptors provided deeper insights influence at on biological
Язык: Английский
Процитировано
0npj Parkinson s Disease, Год журнала: 2025, Номер 11(1)
Опубликована: Март 31, 2025
Abstract Parkinson’s disease (PD) is a progressive neurodegenerative disorder with no effective treatment. Advances in neuroscience and systems biomedicine now enable the use of complex patient-specific vitro models cutting-edge computational tools for data integration, enhancing our understanding PD mechanisms. To explore common biomedical features across monogenic forms, we developed knowledge graph (KG) by integrating previously published high-content imaging RNA sequencing midbrain organoids harbouring LRRK2-G2019S, SNCA triplication, GBA-N370S or MIRO1-R272Q mutations publicly available biological data. Furthermore, generated single-cell dataset derived from idiopathic patients (IPD) to stratify IPD within spectrum forms PD. Despite high degree heterogeneity, found that transcriptomic dysregulation reflected glial cells patient organoids. In addition, ROBO signalling might be involved shared pathophysiology between cases.
Язык: Английский
Процитировано
0BioData Mining, Год журнала: 2025, Номер 18(1)
Опубликована: Апрель 15, 2025
Digital twins in healthcare offer an innovative approach to precision diagnosis, prognosis, and treatment. SynTwin, a novel computational methodology generate digital using synthetic data network science, has previously shown promise for improving prediction of breast cancer mortality. In this study, we validate SynTwin population-level different types from the Surveillance, Epidemiology, End Results (SEER) program National Cancer Institute (USA). We assess its predictive accuracy across varying sample sizes (n = 1,000 30,000 records), mortality rates (35% 60%), study designs, revealing insights into strengths limitations derived prediction. also evaluate effect size 70,000 records) on selected cancers (non-Hodgkin lymphoma, bladder, colorectal cancers). Our results indicate that larger datasets > 10,000) including nearest neighbor model significantly improves performance compared real patients alone. Specifically, AUROCs ranged 0.828 0.884 such as cervix uteri ovarian with twins, 0.720 0.858 when patient data. Similarly, among three cancers, exceeded alone by at least 0.06 narrowing variance increased. These highlight benefit network-based while emphasizing importance considering effective developing models like SynTwin.
Язык: Английский
Процитировано
0Journal of Parkinson s Disease, Год журнала: 2025, Номер unknown
Опубликована: Апрель 27, 2025
Recent years have seen successes in symptomatic drugs for Parkinson's disease, but the development of treatments stopping disease progression continues to fail clinical drug trials, largely due lack efficacy drugs. This may be related limited understanding mechanisms, data heterogeneity, poor target screening and candidate selection, challenges determining optimal dosage levels, reliance on animal models, insufficient patient participation, adherence trials. Most recent applications digital health technologies artificial intelligence (AI)-based tools focused mainly stages before used AI-based algorithms or models discover novel targets, inhibitors indications, recommend candidates dosage, promote remote collection. paper reviews state literature highlights strengths limitations approaches discovery from 2021 2024, offers recommendations future research practice success
Язык: Английский
Процитировано
0Computers in Biology and Medicine, Год журнала: 2025, Номер 192, С. 110285 - 110285
Опубликована: Апрель 29, 2025
To construct an Alzheimer's Disease Knowledge Graph (ADKG) by extracting and integrating relationships among disease (AD), genes, variants, chemicals, drugs, other diseases from biomedical literature, aiming to identify existing treatments, potential targets, diagnostic methods for AD. We annotated 800 PubMed abstracts (ADERC corpus) with 20,886 entities 4935 relationships, augmented via GPT-4. A SpERT model (SciBERT-based) trained on this data extracted relations abstracts, supported databases entity linking refined abbreviation resolution/string matching. The resulting knowledge graph embedding models predict novel relationships. ADKG's utility was validated it UK Biobank predictive modeling. ADKG contained 3,199,276 mentions 633,733 triplets, >5K unique capturing complex AD-related interactions. Its produced evidence-supported predictions, enabling testable hypotheses. In modeling, ADKG-enhanced achieved higher AUROC of 0.928 comparing 0.903 without enhancement. By synthesizing literature-derived insights into a computable framework, bridges molecular mechanisms clinical phenotypes, advancing precision medicine in research. structured underscore its accelerate therapeutic discovery risk stratification.
Язык: Английский
Процитировано
0Frontiers in Neuroinformatics, Год журнала: 2025, Номер 19
Опубликована: Май 2, 2025
Introduction Alzheimer’s disease is a progressive neurodegenerative disorder challenging early diagnosis and treatment. Recent advancements in deep learning algorithms applied to multimodal brain imaging offer promising solutions for improving diagnostic accuracy predicting progression. Method This narrative review synthesizes current literature on applications using neuroimaging. The process involved comprehensive search of relevant databases (PubMed, Embase, Google Scholar ClinicalTrials.gov ), selection pertinent studies, critical analysis findings. We employed best-evidence approach, prioritizing high-quality studies identifying consistent patterns across the literature. Results Deep architectures, including convolutional neural networks, recurrent transformer-based models, have shown remarkable potential analyzing neuroimaging data. These models can effectively structural functional modalities, extracting features associated with pathology. Integration multiple modalities has demonstrated improved compared single-modality approaches. also promise predictive modeling, biomarkers forecasting Discussion While approaches show great potential, several challenges remain. Data heterogeneity, small sample sizes, limited generalizability diverse populations are significant hurdles. clinical translation these requires careful consideration interpretability, transparency, ethical implications. future AI neurodiagnostics looks promising, personalized treatment strategies.
Язык: Английский
Процитировано
0Brain Sciences, Год журнала: 2025, Номер 15(5), С. 523 - 523
Опубликована: Май 19, 2025
This review paper synthesizes the application of knowledge graphs (KGs) in Alzheimer’s disease (AD) research, based on two basic questions, as follows: what types input data are available to construct these graphs, and purpose graph is intended fulfill. We synthesize results from existing works illustrate how diverse structures behave different availability settings with distinct targets AD research. By comparative analysis, we define best methodology practices by type (literature, structured databases, neuroimaging, clinical records) interest (drug repurposing, classification, mechanism discovery, decision support). From this recommend AD-KG 2.0, which a new framework that coalesces into unifying architecture well-defined pathways for implementation. Our key contributions (1) dynamic adaptation adapts methodological elements automatically according both objectives, (2) specialized semantic alignment layer harmonizes terminologies across biological scales, (3) multi-constraint optimization approach building. The accommodates variety applications, including drug patient stratification precision medicine, progression modeling, support. system, tree pipeline layered architecture, offers research precise directions use aligning choice decisions respective goals. provide component designs processes deliver optimal performance varying settings. conclude addressing implementation challenges future translating technologies tool use, specific focus interpretability, workflow integration, regulatory matters.
Язык: Английский
Процитировано
0