SEARCH · PubMed Health

Results for “genomic language model”

Explore indexed PubMed citations for clinical trials, systematic reviews and public health research. Read source abstracts and follow each citation to its original PubMed record.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 55 records · Page 3Linked to original sources

PubMed2006

The use of concept maps during knowledge elicitation in ontology development processes--the nutrigenomics use case.

BACKGROUND: Incorporation of ontologies into annotations has enabled 'semantic integration' of complex data, making explicit the knowledge within a certain field. One of the major bottlenecks in developing bio-ontologies is the lack of a unified methodology. Different methodologies have been proposed for different scenarios, but there is no agreed-upon standard methodology for building ontologies. The involvement of geographically distributed domain experts, the need for domain experts to lead the design process, the application of the ontologies and the life cycles of bio-ontologies are amongst the features not considered by previously proposed methodologies. RESULTS: Here, we present a methodology for developing ontologies within the biological domain. We describe our scenario, competency questions, results and milestones for each methodological stage. We introduce the use of concept maps during knowledge acquisition phases as a feasible transition between domain expert and knowledge engineer. CONCLUSION: The contributions of this paper are the thorough description of the steps we suggest when building an ontology, example use of concept maps, consideration of applicability to the development of lower-level ontologies and application to decentralised environments. We have found that within our scenario conceptual maps played an important role in the development process.

Algorithms↗

PubMed2002

A major susceptibility locus for specific language impairment is located on 13q21.

Children who fail to develop language normally-in the absence of explanatory factors such as neurological disorders, hearing impairment, or lack of adequate opportunity-are clinically described as having specific language impairment (SLI). SLI has a prevalence of approximately 7% in children entering school and is associated with later difficulties in learning to read. Research indicates that genetic factors are important in the etiology of SLI. Studies have consistently demonstrated that SLI aggregates in families. Increased monozygotic versus dizygotic twin concordance rates indicate that heredity, not just shared environment, is the cause of the familial clustering. We have collected five pedigrees of Celtic ancestry that segregate SLI, and we have conducted genomewide categorical linkage analysis, using model-based LOD score techniques. Analysis was conducted under both dominant and recessive models by use of three phenotypic classifications: clinical diagnosis, language impairment (spoken language quotient <85) and reading discrepancy (nonverbal IQ minus non-word reading >15). Chromosome 13 yielded a maximum multipoint LOD score of 3.92 under the recessive reading discrepancy model. Simulation to correct for multiple models and multiple phenotypes indicated that the genomewide empirical P value is <.01. As an alternative measure, we also computed the posterior probability of linkage (PPL), obtaining a PPL of 53% in the same region. One other genomic region yielded suggestive results on chromosome 2 (multipoint LOD score 2.86, genomic P value <.06 under the recessive language impairment model). Our findings underscore the utility of traditional LOD-score-based methods in finding genes for complex diseases, specifically, SLI.

Autistic Disorder↗

PubMed2003

Predicting bacterial transcription units using sequence and expression data.

MOTIVATION: A key aspect of elucidating gene regulation in bacterial genomes is identifying the basic units of transcription. We present a method, based on probabilistic language models, that we apply to predict operons, promoters and terminators in the genome of Escherichia coli K-12. Our approach has two key properties: (i) it provides a coherent set of predictions for related regulatory elements of various types and (ii) it takes advantage of both DNA sequence and gene expression data, including expression measurements from inter-genic probes. RESULTS: Our experimental results show that we are able to predict operons and localize promoters and terminators with high accuracy. Moreover, our models that use both sequence and expression data are more accurate than those that use only one of these two data sources.

Algorithms↗

PubMed2026

Benchmarking large language models for extracting biobank-derived insights into health and disease.

Biobank-scale datasets such as the UK Biobank have become foundational resources for advancing biomedical discovery. Yet the complexity and heterogeneity of these resources, spanning genomics, imaging, clinical records, and metadata, pose substantial barriers to access and interpretation. Large Language Models (LLMs) offer a promising avenue for making such datasets more navigable through natural language interfaces. However, the extent to which current general-purpose LLMs can retrieve and synthesize biobank-specific insights has not yet been systematically evaluated. In this study, we present a reproducible, multi-metric evaluation framework to benchmark the capabilities of leading LLMs. We evaluated six leading large language models: Gemini 3 Pro, Claude Opus 4.5, Claude Sonnet 4.5, GPT-5.2, Mistral Large 2, and DeepSeek V3, on four benchmark tasks designed to assess biobank-related knowledge retrieval. We evaluate model performance across six dimensions (semantic accuracy, factual correctness, domain knowledge, reasoning quality, response depth, and biobank specificity) and assessed output consistency using curated UK Biobank references and a robust random baseline. All models outperformed the baseline by 2× to 3× , with strong statistical separation (p < 0.001), confirming meaningful biobank-specific knowledge retrieval. Gemini 3 Pro achieved the highest overall accuracy across tasks such as keyword synthesis, institution recognition, and topic inference, while Claude Sonnet 4.5 demonstrated the most uniform performance across evaluation dimensions. Our benchmark provides a rigorous framework for evaluating LLMs in biomedical settings. Using the UK Biobank as a real-world testbed, we highlight both the capabilities and limitations of current models, measuring their capacity to recall structured biomedical knowledge consistent with authoritative biobank metadata.

Large Language Models↗

PubMed2026

Decoding gene regulation in plant genomes with artificial intelligence.

One of the central goals of plant functional genomics is to uncover regulatory mechanisms that shape agriculturally important traits to inform crop improvement. Recent advances in machine learning (ML) and artificial intelligence (AI), especially Large Language Models (LLMs), have greatly transformed our ability to derive regulatory information from complex genomics data. This review starts with a brief introduction of recent advances in AI and ML. We then present a plant-focused synthesis of emerging applications of AI- and LLM tools to: (i) predict epigenomic features, regulatory DNA elements, and gene expressions; (ii) infer gene regulatory network; and (iii) estimate post-transcriptional regulation.

Artificial intelligence↗

PubMed2026

Deep generative models in biological sequence and structure analysis and design.

Deep generative models have transformed biological sequence modeling from predictive analysis toward increasingly controllable design. Early biological applications of Variational Autoencoders (VAEs) and Generative Adversarial Networks (GANs) established latent representation learning and sequence synthesis, while recent advances in transformer-based language models, discrete diffusion, flow-matching, and multimodal generative frameworks have substantially expanded the scope of biological design. This review examines generative models for DNA, RNA, and protein sequence design, emphasizing how different model classes represent biological constraints, operate over discrete and continuous spaces, and integrate sequence, structure, and function. We compare VAEs, GANs, autoregressive and masked language models, diffusion models, and flow-based approaches across genomics, transcriptomics, and proteomics, with particular attention to controllability, long-range dependency modeling, structural grounding, generalization, and experimental utility. We further examine evaluation strategies, out-of-distribution generalization, and closed-loop design-build-test-learn workflows that connect in silico generation with empirical validation. We distinguish fundamental modality-dependent constraints including sequence discreteness, context length, structural coupling, and physical or thermodynamic requirements from architecture-dependent advantages that reflect the current state of the field. Current studies suggest that long-context models are particularly useful for genome-scale representation and sequence modeling, whereas structure-aware diffusion, flow-based, and inverse-folding approaches provide better frameworks for geometry-constrained RNA and protein design. This perspective provides a critical framework for understanding the present capabilities, limitations, and convergence of generative approaches toward reliable and experimentally grounded biological design.

Biological sequence analysis↗

PubMed2005

PentaPlot: a software tool for the illustration of genome mosaicism.

BACKGROUND: Dekapentagonal maps depict the phylogenetic relationships of five genomes in a visually appealing diagram and can be viewed as an alternative to a single evolutionary consensus tree. In particular, the generated maps focus attention on those gene families that significantly deviate from the consensus or plurality phylogeny. PentaPlot is a software tool that computes such dekapentagonal maps given an appropriate probability support matrix. RESULTS: The visualization with dekapentagonal maps critically depends on the optimal layout of unrooted tree topologies representing different evolutionary relationships among five organisms along the vertices of the dekapentagon. This is a difficult optimization problem given the large number of possible layouts. At its core our tool utilizes a genetic algorithm with demes and a local search strategy to search for the optimal layout. The hybrid genetic algorithm performs satisfactorily even in those cases where the chosen genomes are so divergent that little phylogenetic information has survived in the individual gene families. CONCLUSION: PentaPlot is being made publicly available as an open source project at http://pentaplot.sourceforge.net.

Algorithms↗

PubMed2005

InterferenceAnalyzer: tools for the analysis and simulation of multi-locus genetic data.

BACKGROUND: Good statistical models for analyzing and simulating multilocus recombination data exist but are not accessible to many biologists because their use requires reasonably sophisticated mathematical and computational implementation. While some labs have direct access to statisticians or programmers competent to carry out such analyses, many labs do not. We have created a platform independent application with an easy-to-use graphical user interface that will carry out such analyses including the simulations needed to bootstrap confidence intervals for the parameters of interest. This software should make multi-locus techniques accessible to labs that previously relied on less powerful and potentially statistically confounded single interval or double interval techniques. RESULTS: We introduce InterferenceAnalyzer, an implementation with a user-friendly graphical interface incorporating previously developed algorithms for the analysis and simulation of multilocus recombination data. We demonstrate the use and features of the program with an example of multilocus tetrad data from the mustard plant, Arabidopsis thaliana, and the yeast, Saccharomyces cerevisiae. CONCLUSION: InterferenceAnalyzer provides easy access to the powerful and appropriate statistical tools for the multi-locus analysis of genetic data.

Algorithms↗

PubMed2005

The use of edge-betweenness clustering to investigate biological function in protein interaction networks.

BACKGROUND: This paper describes an automated method for finding clusters of interconnected proteins in protein interaction networks and retrieving protein annotations associated with these clusters. RESULTS: Protein interaction graphs were separated into subgraphs of interconnected proteins, using the JUNG implementation of Girvan and Newman's Edge-Betweenness algorithm. Functions were sought for these subgraphs by detecting significant correlations with the distribution of Gene Ontology terms which had been used to annotate the proteins within each cluster. The method was implemented using freely available software (JUNG and the R statistical package). Protein clusters with significant correlations to functional annotations could be identified and included groups of proteins know to cooperate in cell metabolism. The method appears to be resilient against the presence of false positive interactions. CONCLUSION: This method provides a useful tool for rapid screening of small to medium size protein interaction datasets.

Algorithms↗

PubMed2006

Using ESTs to improve the accuracy of de novo gene prediction.

BACKGROUND: ESTs are a tremendous resource for determining the exon-intron structures of genes, but even extensive EST sequencing tends to leave many exons and genes untouched. Gene prediction systems based exclusively on EST alignments miss these exons and genes, leading to poor sensitivity. De novo gene prediction systems, which ignore ESTs in favor of genomic sequence, can predict such "untouched" exons, but they are less accurate when predicting exons to which ESTs align. TWINSCAN is the most accurate de novo gene finder available for nematodes and N-SCAN is the most accurate for mammals, as measured by exact CDS gene prediction and exact exon prediction. RESULTS: TWINSCAN_EST is a new system that successfully combines EST alignments with TWINSCAN. On the whole C. elegans genome TWINSCAN_EST shows 14% improvement in sensitivity and 13% in specificity in predicting exact gene structures compared to TWINSCAN without EST alignments. Not only are the structures revealed by EST alignments predicted correctly, but these also constrain the predictions without alignments, improving their accuracy. For the human genome, we used the same approach with N-SCAN, creating N-SCAN_EST. On the whole genome, N-SCAN_EST produced a 6% improvement in sensitivity and 1% in specificity of exact gene structure predictions compared to N-SCAN. CONCLUSION: TWINSCAN_EST and N-SCAN_EST are more accurate than TWINSCAN and N-SCAN, while retaining their ability to discover novel genes to which no ESTs align. Thus, we recommend using the EST versions of these programs to annotate any genome for which EST information is available.TWINSCAN_EST and N-SCAN_EST are part of the TWINSCAN open source software package http://genes.cse.wustl.edu/distribution/download_TS.html.

Algorithms↗

PubMed2005

Effect of dataset selection on the topological interpretation of protein interaction networks.

BACKGROUND: Studies of the yeast protein interaction network have revealed distinct correlations between the connectivity of individual proteins within the network and the average connectivity of their neighbours. Although a number of biological mechanisms have been proposed to account for these findings, the significance and influence of the specific datasets included in these studies has not been appreciated adequately. RESULTS: We show how the use of different interaction data sets, such as those resulting from high-throughput or small-scale studies, and different modelling methodologies for the derivation pair-wise protein interactions, can dramatically change the topology of these networks. Furthermore, we show that some of the previously reported features identified in these networks may simply be the result of experimental or methodological errors and biases. CONCLUSION: When performing network-based studies, it is essential to define what is meant by the term "interaction" and this must be taken into account when interpreting the topologies of the networks generated. Consideration must be given to the type of data included and appropriate controls that take into account the idiosyncrasies of the data must be selected.

Computational Biology↗

PubMed2025

Evaluating Language Models for Biomedical Fact-Checking: A Benchmark Dataset for Cancer Variant Interpretation Verification.

Accurate interpretation of genomic variants is critical for precision oncology but remains slow and dependent on specialized expertise. Public knowledgebases such as the Clinical Interpretation of Variants in Cancer (CIViC) help by curating literature-backed variant interpretations in a structured form, yet verification and review have become major bottlenecks. To address this, we developed CIViC-Fact, a benchmark dataset and pipeline for testing automated systems that verify the accuracy of cancer variant claims. CIViC-Fact links structured claims to sentence-level supporting or refuting evidence from full-text articles, and includes expert annotations and explanations. We evaluated multiple language models. Proprietary models performed well without training, but a smaller open-source model, fine-tuned on CIViC-Fact, achieved the highest accuracy (89%). Applying our fact-checking pipeline to real CIViC entries showed that reviewing less than 20% of content, focusing on flagged entries, would be sufficient to catch over half of all errors. This AI-assisted triage greatly accelerates the review process without replacing or reducing expert insight, ensuring that existing careful oversight remains in place while curators can work more efficiently. CIViC-Fact provides a realistic, high-consequence framework for biomedical fact-checking and a path toward more rigorous and efficient knowledgebase curation.

Journal Article↗

PubMed2026

ELISA (Embedding-Linked Interactive Single-cell Agent): an interpretable hybrid generative Artificial Intelligence agent for expression-grounded discovery in single-cell genomics.

Translating single-cell RNA sequencing (scRNA-seq) data into mechanistic biological hypotheses remains a critical bottleneck, as agentic AI systems lack direct access to transcriptomic representations while expression foundation models remain opaque to natural language. Here, we introduce ELISA (Embedding-Linked Interactive Single-cell Agent), an interpretable framework that unifies single-cell generative pretrained transformer expression embeddings with biomedical bidirectional encoder representations from transformers-based semantic retrieval and large-language model (LLM)-mediated interpretation for interactive single-cell discovery. An automatic query classifier routes inputs to gene marker scoring, semantic matching, or reciprocal rank fusion pipelines depending on whether the query is a gene signature, natural language concept, or mixture of both. Integrated analytical modules perform pathway activity scoring across 60+ gene sets, ligand-receptor interaction prediction using 280+ curated pairs, condition-aware comparative analysis, and cell-type proportion estimation, all operating directly on embedded data without access to the original count matrix. Benchmarked across six diverse scRNA-seq datasets spanning inflammatory lung disease, pediatric and adult cancers, organoid models, healthy tissue, and neurodevelopment, ELISA significantly outperforms CellWhisperer, a classical lexical retriever (BM25), and a random baseline in cell type retrieval (combined permutation test, $p < 2\times 10^{-5}$ for each), with particularly large gains on gene-signature queries (Cohen's $d = 5.98$ for mean reciprocal rank). ELISA replicates published biological findings (mean composite score 0.88), and generates candidate hypotheses through grounded LLM reasoning, bridging the gap between transcriptomic data exploration and biological discovery.

Generative Artificial Intelligence↗

PubMed2006

A Hidden Markov model web application for analysing bacterial genomotyping DNA microarray experiments.

Whole genome DNA microarray genomotyping experiments compare the gene content of different species or strains of bacteria. A statistical approach to analysing the results of these experiments was developed, based on a Hidden Markov model (HMM), which takes adjacency of genes along the genome into account when calling genes present or absent. The model was implemented in the statistical language R and applied to three datasets. The method is numerically stable with good convergence properties. Error rates are reduced compared with approaches that ignore spatial information. Moreover, the HMM circumvents a problem encountered in a conventional analysis: determining the cut-off value to use to classify a gene as absent. An Apache Struts web interface for the R script was created for the benefit of users unfamiliar with R. The application may be found at http://hmmgd.cryst.bbk.ac.uk/hmmgd. The source code illustrating how to run R scripts from an Apache Struts-based web application is available from the corresponding author on request. The application is also available for local installation if required.

Algorithms↗

PubMed2026

A novel transformer model of protein domains for viral taxonomy classification.

MOTIVATION: Viruses with carefully curated taxonomic assignments (such as those in the ICTV taxonomy) still represent only a small fraction of viruses identified through sequencing data from virome or microbiome projects. It is therefore critical to develop methods that can assign viruses at multiple taxonomic ranks, so that a virus deemed novel at a given rank may still be placed into a higher-level taxon. Sequence-similarity-based approaches can classify viruses that share substantial genomic similarity with known viruses (e.g. those belonging to the same species or genus); however, their performance drops significantly when applied to more divergent viruses. Recent deep learning models, such as ViTax, which utilize DNA language models, aim to address these limitations, but their performance also degrades when applied to novel viruses lacking genus-level similarity to known references. Proteins are more conserved than genomic sequences, and the multiple proteins encoded by a virus can be leveraged to reveal evolutionary relationships among viruses. RESULTS: We propose a new tool, D2T (Domain-to-Taxonomy), that leverages recent advances in protein language models to improve viral taxonomic assignment. D2T represents a virus as a sequence of protein domain tokens and learns a transformer-based model for taxonomic classification. Experiments on multiple closed-set and open-set datasets show that D2T excels at assigning higher-level taxonomic labels (family and above). Furthermore, by combining D2T with Kraken2, which performs well at the genus level, the hybrid method (K+D2T) achieves accurate viral taxonomic classification across multiple taxonomic ranks. AVAILABILITY AND IMPLEMENTATION: D2T is available as a GitHub repository at https://github.com/mgtools/D2T.

Viruses↗

PubMed2026

PubMind: literature-based genetic variant extraction and functional annotation using large language models.

Biomedical literature contains extensive functional knowledge on genetic variants, but much remains inaccessible in unstructured text. Existing resources such as ClinVar and HGMD remain limited by coverage, submission bias, update frequency, and sparse annotation. We develop PubMind, an artificial intelligence (AI) framework that uses large language models (LLMs) to triage and extract variant-function-disease associations and supporting evidence from biomedical text. PubMind captures single-nucleotide, copy-number, structural, and gene-fusion variants, and normalizes records to genomic and transcriptomic coordinates. Benchmarking shows >90% accuracy for variant recognition and 99% precision for disease extraction. Applied to >41 million PubMed abstracts and >5 million full-text articles, PubMind generates PubMind-DB, a database of ~1.3 million unique variants with contextual annotations, accessible via web interface and API. Only ~10% of PubMind variants overlap with ClinVar, and >80% of them show concordant pathogenicity labels. PubMind transforms unstructured biomedical text into structured genomic knowledge, advancing variant interpretation for precision medicine.

Large Language Models↗

PubMed2026

Genome- and peak-informed two-stage framework for scATAC-seq cell type identification.

MOTIVATION: Accurate cell type annotation is essential in scATAC-seq analysis, as it underpins the characterization of cellular heterogeneity, the identification of regulatory elements, and downstream biological discovery. However, current annotation methods still face major challenges. First, although some approaches attempt to integrate genomic sequence information, they typically rely on shallow sequence representations and thus fail to capture the long-range dependencies and regulatory signals encoded in DNA. Second, substantial batch effects introduced by different platforms, sequencing batches, or tissue sources remain insufficiently addressed. Existing models often lack robust distribution alignment and domain generalization capabilities, leading to confounding non-biological variation and reduced annotation accuracy across datasets. RESULTS: To overcome these limitations, we propose seqAlignATAC, a two-stage intra-modality annotation framework that integrates sequence-derived embeddings with domain adaptation. In the first stage, we employ a large-scale pretrained nucleotide language model to extract low-dimensional, biologically informative representations from the genomic sequences of chromatin-accessible peaks. In the second stage, these embeddings are fed into a supervised neural network equipped with an adaptive alignment module to mitigate batch effects and harmonize feature distributions between labeled reference and unlabeled target datasets. Extensive experiments across multiple settings demonstrate that seqAlignATAC achieves competitive accuracy and robustness, effectively leveraging genome-level information while alleviating batch-induced distributional discrepancies. AVAILABILITY AND IMPLEMENTATION: The source code of seqAlignATAC is available at: https://github.com/BioCS-Lab/seqAlignATAC.

Humans↗

PubMed2005

Automatic detection of subsystem/pathway variants in genome analysis.

MOTIVATION: Proteins work together in pathways and networks, collectively comprising the cellular machinery. A subsystem (a generalization of pathway concept) is a group of related functional roles (such as enzymes) jointly involved in a specific aspect of the cellular machinery. Subsystems provide a natural framework for comparative genome analysis and functional annotation. A subsystem may be implemented in a number of different functional variants in individual species. In order to reliably project functional assignments across multiple genomes, we have to be able to identify the variants implemented in each genome. The analysis of such variants across diverse species is an interesting problem by itself and may provide new evolutionary insights. However, no computational techniques are presently available for an automated detection and analysis of subsystem variants. RESULTS: Here we formulate the subsystem variant detection problem as finding the minimum number of subgraphs of a subsystem, which is represented as a graph, and solve the optimization problem by integer programming approach. The performance of our method was tested on subsystems encoded in the SEED, a genomic integration platform developed by the Fellowship for Interpretation of Genomes as a component of a large-scale effort on comparative analysis and annotation of multiple diverse genomes. Here we illustrate the results obtained for two expert-encoded subsystems of the biosynthesis of Coenzyme A and FMN/FAD cofactors. Applications of variant detection, to support genomic annotations and to assess divergence of species, are briefly discussed in the context of these universally conserved and essential metabolic subsystems. SUPPLEMENTARY INFORMATION: The details of the variant detection results are available at http://ffas.burnham.org/svar/supp.html.

Animals↗