Search PubMed⌕ Search

SEARCH · Search PubMed

Results for “Protein language model”

Search indexed PubMed citations on genomics, clinical trials, systematic reviews and public health. Explore titles, authors and supplied subject terms, then open the PubMed record.

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

At least 253 records · Page 14Linked to original sources

Extraintestinal considerations in inflammatory bowel disease.

If one reviews the literature with zeal, it is increasingly apparent that few organs escape recruitment when IBD is chronic or progressive. Insights into mucosal pathophysiology have helped with understanding the more frequent extraintestinal manifestations, but the mechanisms attendant to the development of less common events (e.g. acute pancreatitis, concurrent gluten sensitive enteropathy, or active pulmonary disease) remain either poorly studied or obscure. It is particularly interesting, however, to read reports of abnormal pulmonary function, generally of the obstructive type, correlated to measurements of abnormal intestinal permeability in patients with either active pulmonary sarcoid or pulmonary involvement in Crohn's disease. It has been further speculated that similarities in the mucosal immune system of the lung and intestine are responsible for evidence of bronchial hyperreactivity in patients with active IBD. Finally, it is important to recognize that extensions of the inflammatory process are not restricted to the development of organ-based events but may be responsible for some of the most frequent systemic abnormalities detected in IBD patients. It is now also well confirmed that the cytokine environment in IBD can support activated coagulation and, in some clinical situations, overt vascular thrombosis. The cerebrovascular complications of IBD are well recognized and range from peripheral venous thrombosis to central stroke syndromes and pseudotumor cerebri. Reports of focal white matter lesions in the brains of patients with IBD or an increased incidence of polyneuropathy may be other clinical examples of regional microvascular clotting. Microvascular injury appears to be more ubiquitously present, with reports ranging from a speculated primary causative role (e.g., granulomatous vasculitis in the mesenteric circulation) to the utility of nailbed vasospasm, in Crohn's disease, as a clinical marker for disease activity. It is also reported that IL-6 suppression of erythropoietin production is a major feature of the chronic anemia seen in active IBD. Moreover, the capacity of peripheral monocytes from active IBD patients to secrete TNF and IL-8 is reported predictive for the degree of therapeutic response from recombinant erythropoietin. These collected observations constitute another excellent example of the symmetry between basic science and clinical utility. It is from the context of applied basic science that many future therapies will arise. Empiricism will lose much of its appeal as clinical observations will be increasingly translated into cellular language. Already in animal models, elemental diets diminish IL-6-related acute inflammatory injury, and reductions in dietary lipid alter the antigenicity of bacteria. Provocatively, in humans, unconfirmed reports have even associated diet therapy with the resolution of uveitis and pyoderma gangrenosum. It is likely that efforts will also be made to induce oral tolerance if specific triggering proteins are discovered or to alter bowel flora if such an arcane area of investigation becomes resurgent.

Anemia↗

Ionic currents in the human serotonin transporter reveal inconsistencies in the alternating access hypothesis.

We have investigated the conduction states of human serotonin transporter (hSERT) using the voltage clamp, cut-open frog oocyte method under different internal and external ionic conditions. Our data indicate discrepancies in the alternating access model of cotransport, which cannot consistently explain substrate transport and electrophysiological data. We are able simultaneously to isolate distinct external and internal binding sites for substrate, which exert different effects upon currents conducted by hSERT, in contradiction to the alternating access model. External binding sites of coupled Na ions are likewise simultaneously accessible from the internal and external face. Although Na and Cl are putatively cotransported, they have opposite effects on the internal face of the transporter. Finally, the internal K ion does not compete with internal 5-hydroxytryptamine for empty transporters. These data can be explained more readily in the language of ion channels, rather than carrier models distinguished by alternating access mechanisms: in a channel model of coupled transport, the currents represent different states of the same permeation path through hSERT and coupling occurs in a common pore.

Animals↗

Routes are trees: the parsing perspective on protein folding.

An important puzzle in structural biology is the question of how proteins are able to fold so quickly into their unique native structures. There is much evidence that protein folding is hierarchic. In that case, folding routes are not linear, but have a tree structure. Trees are commonly used to represent the grammatical structure of natural language sentences, and chart parsing algorithms efficiently search the space of all possible trees for a given input string. Here we show that one such method, the CKY algorithm, can be useful both for providing novel insight into the physical protein folding process, and for computational protein structure prediction. As proof of concept, we apply this algorithm to the HP lattice model of proteins. Our algorithm identifies all direct folding route trees to the native state and allows us to construct a simple model of the folding process. Despite its simplicity, our model provides an account for the fact that folding rates depend only on the topology of the native state but not on sequence composition.

Algorithms↗

Determination of maternal body composition in pregnancy and its relevance to perinatal outcomes.

Three models and 10 specific methods for determining maternal body composition are discussed and their perinatal relevance reviewed. English language publications (1950 to January 2004) were searched electronically and by hand. Search terms included "body composition," "human," " pregnancy," "obesity," "adiposity," "regional," "2-, 3-, 4-component," "truncal," "peripheral," "central," "visceral" along with specific techniques and outcomes listed subsequently. Three models of body composition are described: 2-component being fat and fat-free mass; 3-component being fat, water, and protein; and 4-component being fat, water, protein, and osseous mineral. Ten techniques of body composition assessment are described: 1) anthropometric techniques including skinfold thicknesses and waist-hip ratio; 2) total body water (isotopically labeled); 3) hydrodensitometry (underwater weighing); 4) air-displacement plethysmography; 5) bio-impedance analysis (BIA); 6) total body potassium (TBK); 7) dual-energy x-ray absorptiometry (DEXA); 8) computed tomography (CT); 9) magnetic resonance imaging (MRI); and 10) ultrasound (USS). Most methods estimate total adiposity. Regional fat distribution-central (truncal) compared with peripheral (limb) or visceral compared with subcutaneous-is important because of regional variation in adipocyte metabolism. Skinfolds, DEXA, CT, MRI, or USS can distinguish central from peripheral fat. CT, MRI, or USS can further subdivide central fat into visceral and subcutaneous. Perinatal outcomes examined in relation to body composition include pregnancy duration, birth weight, congenital anomalies, gestational diabetes, gestational hypertension, and the fetal origins of adult disease. A few studies suggest that central compared with peripheral fat correlates better with birth weight, gestational carbohydrate intolerance, and hypertension. Means of accurately assessing maternal body composition remain cumbersome and impractical, but may more accurately predict perinatal outcomes than traditional assessments such as maternal weight.

Absorptiometry, Photon↗

Agentomics: an agentic system that autonomously develops novel state-of-the-art solutions for biomedical machine learning tasks.

MOTIVATION: Extracting knowledge from biomedical data is crucial for advancing our understanding of biological systems and developing novel therapeutics. The quantity, quality, and resolution of biomedical data constantly evolves, requiring the automation of biomedical machine learning (ML). Existing Automated ML tools lack flexibility, while large language models (LLMs) struggle to consistently deliver reproducible machine learning codebases, and existing LLM Agent-powered solutions lag behind human-engineered ML models. RESULTS: Here, we introduce Agentomics, an autonomous LLM-powered agentic system for end-to-end ML experimentation. Given a biomedical dataset, Agentomics implements various ML modeling strategies, and produces a ready-to-use ML model. Agentomics introduces strict validation checkpoints for standard ML development steps, allowing gradual development on top of working code with defined interfaces and validated artifacts. Further, it offers native support for biomedical foundation models that can be leveraged during experimentation. The generic nature of Agentomics allows the user to create ML solutions for a large variety of datasets and use various LLMs. We evaluate Agentomics across 20 datasets from the domains of Protein Engineering, Drug Discovery, and Regulatory Genomics. When benchmarked against other agentic systems, Agentomics outperformed them in all tested domains. When benchmarked against human expert solutions, Agentomics generated novel state-of-the-art models for 11/20 established benchmark datasets. AVAILABILITY AND IMPLEMENTATION: Agentomics is implemented in Python. Source code and documentation are freely available at: https://github.com/BioGeMT/Agentomics-ML.

Machine Learning↗

Experimental and clinical applications of molecular cell biology in nutrition and metabolism.

Rapid advances in molecular biology have yielded important new techniques for understanding the cellular mechanisms of normal homeostasis and disease. In particular, molecular laboratory methodologies have become important investigative tools for nutritional studies. Detection techniques for specific DNAs, RNAs, and proteins allow direct examination of cellular regulation of protein expression in health and illness. Construction of transgeneic models by recent techniques of inserting foreign genes into experimental animals has provided novel models for studies of cellular metabolism. In addition, molecular biology has had impact on clinical nutrition and therapy. Molecular techniques not only allow for early diagnosis of many inborn genetic errors of metabolism, recombinant technology has also provided for large-scale production of proteins and hormones of potential therapeutic value. The possibility for direct gene therapies is also nearing reality. Hence, understanding the language of molecular biology and the recent developments in this field is not only of research interest, but is also of clinical relevance.

Animals↗

CellML2SBML: conversion of CellML into SBML.

UNLABELLED: CellML and SBML are XML-based languages for storage and exchange of molecular biological and physiological reaction models. They use very similar subsets of MathML to specify the mathematical aspects of the models. CellML2SBML is implemented as a suite of XSLT stylesheets that, when applied consecutively, convert models expressed in CellML into SBML without significant loss of information. The converter is based on the most recent stable versions of the languages (CellML version 1.1; SBML Level 2 Version 1), and the XSLT used in the stylesheets adheres to the XSLT version 1.0 specification. Of all 306 models in the CellML repository in April 2005, CellML2SBML converted 91% automatically into SBML. Minor manual changes to the unit definitions in the originals raised the percentage of successful conversions to 96%. AVAILABILITY: http://sbml.org/software/cellml2sbml/. SUPPLEMENTARY INFORMATION: Instructions for use and further documentation available on http://sbml.org/software/cellml2sbml/

Cell Physiological Phenomena↗

Design of highly functional genome editors by modelling CRISPR-Cas sequences.

Gene editing has the potential to solve fundamental challenges in agriculture, biotechnology and human health. CRISPR-based gene editors derived from microorganisms, although powerful, often show notable functional tradeoffs when ported into non-native environments, such as human cells1. Artificial-intelligence-enabled design provides a powerful alternative with the potential to bypass evolutionary constraints and generate editors with optimal properties. Here, using large language models2 trained on biological diversity at scale, we demonstrate successful precision editing of the human genome with a programmable gene editor designed with artificial intelligence. To achieve this goal, we curated a dataset of more than 1 million CRISPR operons through systematic mining of 26 terabases of assembled genomes and metagenomes. We demonstrate the capacity of our models by generating 4.8× the number of protein clusters across CRISPR-Cas families found in nature and tailoring single-guide RNA sequences for Cas9-like effector proteins. Several of the generated gene editors show comparable or improved activity and specificity relative to SpCas9, the prototypical gene editing effector, while being 400 mutations away in sequence. Finally, we demonstrate that an artificial-intelligence-generated gene editor, denoted as OpenCRISPR-1, exhibits compatibility with base editing. We release OpenCRISPR-1 to facilitate broad, ethical use across research and commercial applications.

CRISPR-Cas Systems↗

Domain fusion analysis by applying relational algebra to protein sequence and domain databases.

BACKGROUND: Domain fusion analysis is a useful method to predict functionally linked proteins that may be involved in direct protein-protein interactions or in the same metabolic or signaling pathway. As separate domain databases like BLOCKS, PROSITE, Pfam, SMART, PRINTS-S, ProDom, TIGRFAMs, and amalgamated domain databases like InterPro continue to grow in size and quality, a computational method to perform domain fusion analysis that leverages on these efforts will become increasingly powerful. RESULTS: This paper proposes a computational method employing relational algebra to find domain fusions in protein sequence databases. The feasibility of this method was illustrated on the SWISS-PROT+TrEMBL sequence database using domain predictions from the Pfam HMM (hidden Markov model) database. We identified 235 and 189 putative functionally linked protein partners in H. sapiens and S. cerevisiae, respectively. From scientific literature, we were able to confirm many of these functional linkages, while the remainder offer testable experimental hypothesis. Results can be viewed at http://calcium.uhnres.utoronto.ca/pi. CONCLUSION: As the analysis can be computed quickly on any relational database that supports standard SQL (structured query language), it can be dynamically updated along with the sequence and domain databases, thereby improving the quality of predictions over time.

Alternative Splicing↗

Tilted peptides: the history.

Nature has selected peptide motifs for protein functions. It is clear that specific sequence motifs can identify families of enzymes. These sequence motifs are one dimensional signatures and nature has also developed two dimension motifs which cannot be read in the one dimension of sequence language but can be detected in the three dimensional properties of a secondary structure. One of such motifs is tilted peptides. They do not correspond to any consensus of sequence but correspond to a consensus motif where hydrophobicity balance is used as a functional device. In the nineteen eighties, the first tilted peptide was deciphered from the sequence of a virus fusion protein by molecular modelling. It was described as a protein fragment hydrophobic enough to insert into a membrane but too short to span it. The fragment exhibited an asymmetric distribution of hydrophobicity along the helix long axis and hence, was unable to lie parallel or perpendicular to a membrane surface but adopted an orientation in between. Hydrophobicity motif was a very new and very challenging concept and tilted peptides were rapidly found to be involved in several mechanisms of virus fusion. They were also found to be involved in protein secretion and future studies could establish their involvement in the destabilization of 3D protein structures and in the alpha to beta transconformations, which drive the generation of amyloid deposits.

Amino Acid Sequence↗

Harnessing the Power of Large Language Models for Drug Discovery: A Systematic Review of Current Applications and Future Directions.

INTRODUCTION: The demand for inventive approaches to drug discovery has increased due to the rising costs, time, and failure rates in pharmaceutical research. Large Language Models (LLMs), with their sophisticated natural language processing and generative capabilities, have become potent instruments that have the potential to revolutionize biomedical research. The function of LLMs in different phases of drug development is methodically examined in this article. METHODS: The PRISMA 2020 principles were adhered to in this systematic study. A thorough search for research published between 2018 and 2025 was done using PubMed, Scopus, Web of Science, and Google Scholar. The search terms "large language model," "transformer," "drug discovery," and important sub-domains (such as "de-novo design" and "ADMET") were merged, and two reviewers independently screened the results. Predetermined inclusion and exclusion criteria were used to filter studies for relevance. 98 studies out of the 1,285 records that were initially retrieved met the requirements for the final qualitative synthesis. RESULTS: 98 studies that demonstrated the use of LLMs in various drug discovery domains were found during the review. These covered molecular generation, genomics, protein-ligand modeling, ADME/T and toxicity profiling, drug-target interaction and DTI prediction, and biomedical text mining. 42 different LLM-based tools were mapped, including BioBERT, SciSpacy, Drug- LLM, DNA-BERT, GPT-4, and ChatGPT. Predictive accuracy, hypothesis creation, target prioritization, and multi-modal data integration all showed notable gains with these techniques. DISCUSSION: By providing scalable, precise, and effective solutions for data-driven drug discovery, LLMs are revolutionizing the pharmaceutical industry. They allow for the creation of hypotheses and individualized insights across multi-modal biological data, and they perform better than conventional approaches in a number of subdomains. Improvements in performance were task-dependent; the most consistent gains occurred for biomedical text mining, disease-genedrug relationship mapping and drug-target interaction prediction tasks. Yet most evidence for clinical applications is still derived from retrospective studies and benchmark datasets, suggesting a higher need for prospective validation. CONCLUSION: There is revolutionary potential in incorporating LLMs into drug discovery processes. Clinical translation and regulatory uptake will depend heavily on collaborative validation, ethical deployment, and standardization as models become more multimodal and interpretable. Before normal use, extensive prospective benchmarking and head-to-head comparisons with established chemoinformatics pipelines are necessary.

De novo design↗

Software tools that facilitate kinetic modelling with large data sets: an example using growth modelling in sugarcane.

A solution to manage cumbersome data sets associated with large modelling projects is described. A kinetic model of sucrose accumulation in sugarcane is used to predict changes in sucrose metabolism with sugarcane internode maturity. This results in large amounts of output data to be analysed. Growth is simulated by reassigning maximal activity values, specific to each internode of the sugarcane plant, to parameter attributes of a model object. From a programming perspective, only one model definition file is required for the simulation software used; however, the amount of input data increases with each extra interrnode that is modelled, and likewise the amount of output data that is generated also increases. To store, manipulate and analyse these data, the modelling was performed from within a spreadsheet. This was made possible by the scripting language Python and the modelling software PySCeS through an embedded Python interpreter available in the Gnumeric spreadsheet program.

Algorithms↗

MOLE: a data management application based on a protein production data model.

MOLE (mining, organizing, and logging experiments) has been developed to meet the growing data management and target tracking needs of molecular biologists and protein crystallographers. The prototype reported here will become a Laboratory Information Management System (LIMS) to help protein scientists manage the large amounts of laboratory data being generated due to the acceleration in proteome research and will furthermore facilitate collaborations between groups based at different sites. To achieve this, MOLE is based on the data model for protein production devised at the European Bioinformatics Institute (Pajon A, et al., Proteins in press).

Algorithms↗

3D molecular graphics on the World Wide Web.

The Virtual Reality Modeling Language (VRML) is a new tool for the information transfer on the World Wide Web (WWW). We present a short description and some applications in the field of molecular graphics of this new, object oriented language. It is shown that the usage of VRML is a very efficient and powerful method for the transport of molecular models over the net.

Binding Sites↗

The influence of early malnutrition on subsequent behavioral development III. Learning disabilities as a sequel to malnutrition.

The academic performance of 129 Barbadian children (77 boys and 52 girls), who were 5-11-yr-old and suffered moderate to severe protein-energy malnutrition in the first year of life, was compared with the performance of matched comparisons, children who had no history of malnutrition. Children with a history of malnutrition were found to have lower performance on eight out of nine academic subject areas, namely, language arts, mathematics, general science, social science, reading, health, religion, and arts/crafts. Socioeconomic factors in the backgrounds of the children were also examined. A model is presented clarifying the interrelationships of malnutrition, socioeconomic status and school performance. Reduced school performance in the previously malnourished children can be largely accounted for by deficits in classroom behavior, and, to a lesser extent by a reduction in I.Q. Current socioeconomic status is not directly involved in altering academic performance whereas the early history of malnutrition and its accompanying conditions at the time of the illness are leading contributors to altered behavioral outcome and school performance.

Analysis of Variance↗

Current status of proteomic standards development.

The generation of proteomic data is becoming ever more high throughput. Both the technologies and experimental designs used to generate and analyze data are becoming increasingly complex. The need for methods by which such data can be accurately described, stored and exchanged between experimenters and data repositories has been recognized. Work by the Proteome Standards Initiative of the Human Proteome Organization has laid the foundation for the development of standards by which experimental design can be described and data exchange facilitated. The Minimum Information About a Proteomic Experiment data model describes both the scope and purpose of a proteomics experiment and encompasses the development of more specific interchange formats such as the mzData model of mass spectrometry. The eXtensible Mark-up Language-MI data interchange format, which allows exchange of molecular interaction data, has already been published and major databases within this field are supplying data downloads in this format.

Databases, Protein↗

A molecular code for splicing silencing: configurations of guanosine-rich motifs.

Alternative pre-mRNA splicing is frequently used to expand the protein-coding capacity of genomes, and to regulate gene expression at the post-transcriptional level. It is a significant challenge to decipher the molecular language of tissue-specific splicing because the inherent flexibility of these mechanisms is specified by numerous short sequence motifs distributed in introns and exons. In the present study, we employ the glutamate NMDA (N-methyl-D-aspartate) R1 receptor (GRIN1) transcript as a model system to identify the molecular determinants for a brain region-specific exon silencing mechanism. We identify a set of guanosine-rich motifs that function co-operatively to regulate the CI cassette exon in a manner consistent with its in vivo splicing pattern. Whereas hnRNP (heterogeneous nuclear ribonucleoprotein) A1 mediates silencing of the CI cassette exon in conjunction with the guanosine-rich motifs, hnRNP H functions as an antagonist to silencing. Genome-wide analysis shows that, while this motif pattern is rarely present in human and mouse exons, those exons for which the pattern is conserved are generally found to be skipped exons. The identification of a similar arrangement of guanosine-rich motifs in transcripts of the hnRNP H family of splicing factors has implications for their co-ordinate regulation at the level of splicing.

Alternative Splicing↗