SEARCH · PubMed Health

Results for “Deep generative models”

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 91 records · Page 5Linked to original sources

PubMed1989

Flexibility plot of proteins.

The flexibility plot of a protein lies on the observation that amino acid residues with the highest turn potential, i.e. located in highly mobile regions of protein surface, also possess the smallest volumes as well as the lowest hydrophobicities. The plot is generated by shifting a five residue window along the protein sequence and calculating the value of the hydrophobicity-volume product for consecutive quintuplets of amino acid residues. The concomitant occurrence of small volumes and low hydrophobicities results in very deep minima. A threshold value has also been introduced in order to discriminate significant minima. To substantiate the interpretation that the selected minima actually indicate very flexible segments of a protein (loops, turns, etc.), we have compared plots obtained for model proteins (lysozyme, myoglobin, ribonuclease, trypsin, thermolysin and T4 lysozyme) with X-ray thermal factors profiles available for the same proteins. When compared to thermal profiles, the majority of flexible segments evidenced by our plots have been found to be in agreement with regions characterized by high thermal factors. Results have also been discussed in the light of local organization possessed by examined proteins.

Amino Acids↗

PubMed1979

Evaluation of electrofulguration in control of bleeding of experimental gastric ulcers.

The safety and efficacy of electrofulguration for control of bleeding from standard canine experimental gastric ulcers was studied. At settings of 2, 5, and 8 on a Valleylab SSE-3 generator, 0.5-sec applications provided effective hemostasis. However, a setting of 2 required an excessive number of applications. Settings of 5 and 8 showed deep injury to the muscularis externa when examined histologically. In an attempt to reduce the depth of injury, a more easily ionizable gas mixture of 50% argon gas and 50% CO2 was compared to CO2 alone. At a generator setting of 5 with 0.5-sec applications the argon-CO2 mixture produced slightly less deep injury than CO2 alone, but the difference was not significant. Although electrofulguration was effective in stopping bleeding in these experiments, the tissue injury was unpredictable and deep.

Animals↗

PubMed2026

Flexible use of conserved motifs constrains genome access in cell type evolution.

Cell types can be organized into related families, but the regulatory mechanisms that define and maintain these families across deep evolutionary time remain unknown. Here, combining single-nucleus multi-omic sequencing with deep learning to analyse the accessible genomes of two groups of vastly divergent animals including flatworms and vertebrates, we find that hundreds of accessibility-dictating sequence motifs partition into distinct yet conserved sets, or 'vocabularies', each associated with a specific cell type family. However, combinatorial relationships among these motifs preferred by individual cell types are largely species specific. Deep-learning models trained on one species accurately predict family-level chromatin accessibility in distantly related species, albeit frequently rely on different motifs from shared vocabularies to reach convergent predictions. By contrast, models trained on individual cell types within a family lose cross-species predictive power, indicating that the regulatory syntax governing cell type-level identity evolves rapidly. We propose a 'collective maintenance' model in which motif vocabularies defining cell type families are evolutionarily stable, while recombination of these motifs generates cell type-specific regulatory programmes. This suggests that family identity is maintained collectively by large, conserved pools of regulatory factors, analogous to the logic of developmental homology, where character identity persists through network-level conservation despite extensive rewiring.

Journal Article↗

PubMedDate not supplied

Optimization of the intensity gain of multiple-focus phased-array heating patterns.

A new technique for enhancing the intensity gain at the focal points in multiple-focus patterns is introduced. The new technique is shown to be effective in reducing the interference typically associated with multiple-focus patterns. This reduction in interference patterns allows multiple-focus scanning to generate highly localized heating. Simulation results indicate that multiple-focus scanning not only provides an alternative to single-focus scanning, but also achieves better localization in the heating pattern. The maximization of intensity gain of multiple-focus heating patterns significantly reduces the pre-focal-depth high-temperature regions that can be caused by single-focus scanning. This is shown by computer simulation of a two-dimensional cylindrical-section array (CSA2D) as a heating applicator. Two series of simulations are presented in which different scan trajectories were used to therapeutically heat a small deep-seated target volume. In every case the heating pattern was generated using single-focus scanning and multiple-focus scanning (with and without intensity gain maximization). Multiple-focus scanning with gain maximization offers the best localization of heating to the target volume of the three methods.

Biophysical Phenomena↗

PubMed2026

Profiler: an open web platform for multi-omics analysis.

MOTIVATION: High-throughput multi-omics technologies produce increasingly large and heterogeneous datasets that are difficult to analyze without advanced computational expertise. Existing bioinformatics tools are often fragmented or limited to specific omics types, hindering reproducibility and accessibility. There is a critical need for an integrated, user-friendly, and scalable platform capable of supporting multi-omics analyses across different data modalities. RESULTS: We present Profiler, an open-source, modular platform that unifies data import, quality control, preprocessing, statistical testing, machine and deep learning, biomarker discovery, pathway and drug-target enrichment, and survival modeling within a single reproducible environment. Built in Python with Streamlit, Profiler is available as both a web-based platform deployed on high-performance computing and a desktop version for local execution, enabling flexible usage across computational infrastructures. Profiler supports diverse omics modalities, including proteomics, transcriptomics, lipidomics, and electroencephalogram data. Through applications to glioblastoma proteomic, pancancer, and multi-omics datasets, Profiler reproduced known molecular subtypes, revealed potential therapeutic targets, and generated fully traceable analysis reports within minutes. By integrating advanced analytics behind an intuitive interface, Profiler democratizes multi-omics analysis and provides a robust, scalable foundation for systems biology and precision medicine research. AVAILABILITY AND IMPLEMENTATION: Profiler is open-source and freely available via its web platform (https://prism-profiler.univ-lille.fr) and GitHub (web version: https://github.com/yanisZirem/Profiler_v1_requests_datatests, desktop version: https://github.com/yanisZirem/prism-profiler), and archived on Zenodo (DOI: https://doi.org/10.5281/zenodo.17478158).

Software↗

PubMed1991

Source modelling of the rolandic focus.

In rolandic epilepsy, consideration of the stereotyped ictal symptomatology suggests that the epileptic zone is likely to be in the same cortical structure in different patients. Routine EEG tracings of the interictal spike activity suggests a deep Sylvian fissure location. On the basis of the predominantly tangential potential field at the peak spike negativity seen in this group of patients, the inferior bank of the Sylvian fissure appears to be a good candidate. Without invasive studies, little refinement to this rather imprecise localization can be made as there is neither neurologic deficit nor lesion to provide a marker on radiological imaging. However, the application of source modelling technique using a simple single-dipole spherical head model has resulted in improved understanding of the generator behaviour, and facilitated the generation of new ways of analyzing spikes (e.g., stability index). Review of newer quantitative approaches including matrix and singular value decomposition of the dataset, spatial-temporal constrained source estimates etc. suggest other fruitful approaches. At least in some patients with partial epilepsy, the source characteristics of interictal scalp spikes appear to contain information of the ictal generator. Under certain circumstances, such derived information which is not otherwise available from routine electrophysiology may influence clinical management and prognosis. This is an additional bonus to the primary objectives of quantification and data reduction.

Brain↗

PubMed2026

Treemble: a graphical tool to generate Newick strings from phylogenetic tree images.

SUMMARY: Phylogenetic trees are ubiquitous and central to biology, but most published trees are available only as visual diagrams and not in the machine-readable Newick format. There are, thus, thousands of published trees in the scientific literature that are unavailable for follow-up analyses, comparisons, and supertree construction. Experts can easily read such diagrams, but the manual construction of a Newick string from a diagram is laborious, error-prone, and time-consuming. Previous attempts to semi-automate the reading of tree images relied on image processing techniques. These often encounter difficulties as typical published tree diagrams contain various graphical elements and annotations that overlap the branches, such as error bars on internal nodes. Here we introduce Treemble, a user-friendly desktop application for generating Newick strings from tree images. The user simply clicks to mark node locations, assisted by a deep learning-based node detection tool, and Treemble algorithmically assembles the tree from the node coordinates alone. Treemble also facilitates the automatic reading of tip name labels and can be used for both rectangular and circular trees. AVAILABILITY AND IMPLEMENTATION: Treemble is a native desktop application for macOS and Windows and is freely available, with documentation, at treemble.org. Source code is available at github.com/John-Allard/Treemble. The trained node detection model is available at huggingface.co/John-Allard/treemble-1.

Phylogeny↗

PubMed1994

A spatiotemporal dipole model of the stimulus preceding negativity (SPN) prior to feedback stimuli.

Ten subjects performed a time production task, in which they were instructed to press a button four seconds after the presentation of an auditory stimulus. Two seconds after the button press they received either auditory or visual feedback on the temporal accuracy of their response. In such a paradigm negative slow brain potentials can be recorded preceding the response (Movement Preceding Negativity, MPN) as well as preceding the feedback stimulus (Stimulus Preceding Negativity, SPN). Spatiotemporal dipole modelling is used to gain insight in the possible generators of MPN and SPN. From the models it follows that the MPN can be described by one contralateral radial dipole and a bilateral pair of tangential dipoles. All three dipoles are located near central electrode positions, so the generators of the MPN probably reside within the motor cortex. The SPN is modelled by a bilateral frontotemporal pair of dipoles, hypothetically representing activation of the Insulae Reili. The insular cortex is involved in the processing of affective-motivational input, such as carried by the feedback in the present paradigm. However, processing of the information content of the feedback stimulus might by itself also activate the frontal cortex. Both the response and the feedback stimulus are followed by a positive peak, which can be described by the same deep posterior dipole. Both peaks probably represent a P3, which is related to context updating.

Adult↗

PubMed1985

Comparative theoretical performance for two types of regional hyperthermia systems.

Regional hyperthermia systems have drawn attention because of their potential for depositing power noninvasively in deep-seated tumors. Two such systems that have received clinical attention because of their ability to deposit significant amounts of power in tissue are magnetic induction devices and annular phased array applicators. In this paper, theoretical calculations for the specific absorption rate (SAR) and the resulting temperature distributions for these systems are compared. The finite element method is used in the formulation of both the electromagnetic and thermal boundary value problems. Six detailed patient models based on CT-scan data from the pelvic, visceral, and thoracic regions are generated to simulate a variety of tumor locations. In general, the annular phased array deposited more power within the tumor and produced better temperature distributions than the magnetic induction device. However, the ratio of the maximum power absorbed by the tumor to the maximum power absorbed in normal tissue does not appear to be high enough for either device to heat significant portions of perfused tumors to therapeutic temperatures under a wide range of physiological conditions. The results contained herein should aid the physician in comparative treatment planning with existing regional hyperthermia systems.

Electromagnetic Fields↗

PubMed2026

Sequence optimization targeting mRNA stability enhances monoclonal antibody titers in CHO cells.

This study presents a DNA sequence optimization approach that integrates mRNA stability as a tunable design parameter to enhance monoclonal antibody expression in Chinese hamster ovary (CHO) cells. A comprehensive combinatorial library of synonymous coding-sequence variants of an IgG1 light chain was integrated as single copies at a defined genomic locus in CHO cells with identical regulatory elements. Steady-state mRNA abundance, quantified by deep sequencing of gDNA and mRNA, served as a proxy for mRNA stability. These data were used to train a machine learning model that predicts mRNA abundance from coding sequence using embeddings from a pre-trained nucleotide transformer. This abundance predictor, together with established translational metrics, was incorporated into a genetic algorithm for multi-objective codon optimization. As proof-of-concept, we optimized sequences encoding Trastuzumab to either maximize or minimize the abundance criterion and obtained benchmark sequences from two commercial providers. Using targeted integration, we generated CHO cell lines and measured protein titer and cell-specific productivity. Sequences optimized for high abundance significantly increased intracellular mRNA levels (+41%), protein titer (+59%), and cell-specific productivity (+85%) relative to low-abundance designs, while viable cell densities remained comparable. Compared to commercial benchmarks, high-abundance sequences achieved significantly higher titer (+70%) and cell-specific productivity (+98%). These findings establish mRNA stability as a practical and complementary design parameter for codon optimization in monoclonal antibody production, with potential applicability to other proteins and expression systems.

CHO↗

PubMed1990

Modelling coronary thrombosis from nonanticoagulated human blood in vitro.

The prevalence of ruptured atheromatous plaques underlying the adherent thrombus in the infarct-related coronary arteries, is well documented. In the thrombotic process associated with plaque rupture, hemodynamic forces and the interaction of platelets with exposed collagen fibers play the decisive roles. The shear-induced hemostasis from a nonanticoagulated human blood sample, perfused through polyethylene tubing, was used to simulate rheological changes in the coronary circulation due to plaque disruption. When the hemodynamic conditions of a plaque fissure were mimicked, the sequence of events corresponded to that in vivo: hemostasis (i.e., platelet plug formation in the wall) resulted in the formation of an occlusive thrombus in the lumen of the tubing. Further, the thrombus growth on a collagen fiber, mounted in the lumen of polyethylene tubing through which nonanticoagulated human blood was perfused, was used to mimick the exposure of thrombogenic elements during deep vessel wall injury and the formation of thrombus superimposed on plaque disruption. Morphology of both types of thrombi revealed large numbers of neutrophils and monocytes associated with the platelet mass. The mechanisms of thrombotic reactions were characterized by antagonists and monoclonal antibodies against the platelet activation pathways. Generation of thrombin at an early stage was shown to be the key event and determinant of the final outcome of both thrombotic reactions. It is suggested that the simultaneous measurements of shear-induced hemostasis, clotting, and platelet-collagen interaction from nonanticoagulated human blood provide close experimental approximations to the pathological process of acute coronary syndromes, namely thrombus formation at the site of a disrupted atherosclerotic plaque.

Adolescent↗

PubMed2026

Conditional Diffusion Model-Based Method for Annotation of Antibiotic Resistance Gene Properties.

The crisis of bacterial antibiotic resistance, which has led to a decline in the effectiveness of antibiotics originally used to combat bacterial infections, has emerged as an urgent challenge for public health. Antibiotic resistance genes (ARGs) are one of the key reasons for bacteria to develop resistance to antibiotics. Therefore, accurately identifying and annotating the critical properties of ARGs is of great importance for addressing the antibiotic resistance emergency. Although existing deep learning models demonstrate remarkable effectiveness in extracting local features from sequence data, they still face limitations in the capacity to further gain the enriched latent representations within the data. To address the critical challenge of extracting higher-quality representations from ARGs sequence data, we propose a novel ARGs properties annotation method based on the conditional diffusion model which is used to learn latent representations through domain-specific knowledge injection. Specifically, during the conditional information integration phase, we systematically incorporate ARGs' domain knowledge to guide the diffusion process in generating high-quality latent representations. To overcome information redundancy caused by direct concatenation of conditional information and intermediate features, we design a cross-attention mechanism that enables feature fusion between heterogeneous information sources, thereby enhancing further the quality of obtained representations. Experimental results on widely used data sets demonstrate the framework's effectiveness in achieving superior prediction performance compared to existing methods.

Anti-Bacterial Agents↗

PubMed1983

Capillary transport of H2 gas generated locally in renal tissue.

Previous measurements by microspheres have shown a higher blood flow in outer cortex and a lower blood flow in inner cortex than found by diffusible tracers. During vasodilation microspheres have indicated a disproportionate increase in deep cortical blood flow, whereas diffusible tracer distributions remained unchanged. These discrepancies could possibly be explained by a variable net inward transport of diffusible tracers in postglomerular vessels, the transport existing in control, but disappearing during vasodilation. To test this hypothesis H2 gas was produced electrolytically for 1 s at a platinum electrode in midcortex and the resulting gas concentration curve measured polarographically at two electrodes placed above and below the source. Analysis of a mathematical model showed that the ratio of the curve maxima at the two electrodes (Cmo/Cmi) would best reveal a radial net transport. Average Cmo/Cmi at 25 positions in 7 clamped dog kidneys was close to unity, but rose to 1.24 at control flow. During acetylcholine infusion Cmo/Cmi rose to 1.68. Local washout rates at the two electrodes increased equally. Calculations indicated a small outwardly directed net transport in control (3 X 10(-4) cm/s), becoming slightly reinforced during vasodilation (5 X 10(-4) cm/s). Thus the control transport direction is opposite to the hypothesis, and the change during vasodilation was estimated to be too small to explain the disparity between diffusible tracer uptake and microsphere distribution in control. H2 concentration maximum was obtained earlier under control flow than in the clamped kidney, indicating an increase in apparent D of the gas in tissue from 3 X 10(-5) cm2/s to 5 X 10(-5) cm2/s, probably due to mixing of H2 gas in the capillary net work.

Acetylcholine↗

PubMed2025

Efficient Detection and Characterization of Targets of Natural Selection Using Transfer Learning.

Natural selection leaves detectable patterns of altered spatial diversity within genomes, and identifying affected regions is crucial for understanding species evolution. Recently, machine learning approaches applied to raw population genomic data have been developed to uncover these adaptive signatures. Convolutional neural networks (CNNs) are particularly effective for this task, as they handle large data arrays while maintaining element correlations. However, shallow CNNs may miss complex patterns due to their limited capacity, while deep CNNs can capture these patterns but require extensive data and computational power. Transfer learning addresses these challenges by utilizing a deep CNN pretrained on a large dataset as a feature extraction tool for downstream classification and evolutionary parameter prediction. This approach reduces extensive training data generation requirements and computational needs while maintaining high performance. In this study, we developed TrIdent, a tool that uses transfer learning to enhance detection of adaptive genomic regions from image representations of multilocus variation. We evaluated TrIdent across various genetic, demographic, and adaptive settings, in addition to unphased data and other confounding factors. TrIdent demonstrated improved detection of adaptive regions compared to recent methods using similar data representations. We further explored model interpretability through class activation maps and adapted TrIdent to infer selection parameters for identified adaptive candidates. Using whole-genome haplotype data from European and African populations, TrIdent effectively recapitulated known sweep candidates and identified novel cancer, and other disease-associated genes as potential sweeps.

Selection, Genetic↗

PubMed1988

Extraction of "deep" components from scalp EEG.

In an attempt to delineate the relative contribution of surface versus deep components in the EEG of patients with 3 per second generalized spike-wave discharges and clinical petit mal seizures, a mathematical method was devised which allows the splitting of the EEG into two major subsystems. It is based on the observation that broad electrical fields tend to represent activity at deeper structures while discrete narrow fields centered at one electrode position tend to be of more superficial origin. Since source derivation intentionally suppresses broad potential fields, a differentiation between superficial and deep activity can be achieved by subtracting the source density values from the corresponding electrode potential values. This will provide those aspects of the EEG which are contributed mainly by deep generators. The resultant data can then be subjected to eigenfunction analysis which yields few uncorrelated components. The percentage of contribution of each electrode to the total component thus derived can then be displayed as a topographic map. When this methodology was applied to ictal EEGs of three patients consistent results were obtained. The "deep" data yielded mainly two components with mutually perpendicular directions.

Cerebral Cortex↗

PubMed2025

Efficient detection and characterization of targets of natural selection using transfer learning.

Natural selection leaves detectable patterns of altered spatial diversity within genomes, and identifying affected regions is crucial for understanding species evolution. Recently, machine learning approaches applied to raw population genomic data have been developed to uncover these adaptive signatures. Convolutional neural networks (CNNs) are particularly effective for this task, as they handle large data arrays while maintaining element correlations. However, shallow CNNs may miss complex patterns due to their limited capacity, while deep CNNs can capture these patterns but require extensive data and computational power. Transfer learning addresses these challenges by utilizing a deep CNN pre-trained on a large dataset as a feature extraction tool for downstream classification and evolutionary parameter prediction. This approach reduces extensive training data generation requirements and computational needs while maintaining high performance. In this study, we developed TrIdent, a tool that uses transfer learning to enhance detection of adaptive genomic regions from image representations of multilocus variation. We evaluated TrIdent across various genetic, demographic, and adaptive settings, in addition to unphased data and other confounding factors. TrIdent demonstrated improved detection of adaptive regions compared to recent methods using similar data representations. We further explored model interpretability through class activation maps and adapted TrIdent to infer selection parameters for identified adaptive candidates. Using whole-genome haplotype data from European and African populations, TrIdent effectively recapitulated known sweep candidates and identified novel cancer, and other disease-associated genes as potential sweeps.

Journal Article↗

PubMed1976

Vitamin-A-induced mucous metaplasia. An in vitro system for modulating tight and gap junction differentiation.

Stratified squamous epithelia from 14-day chick embryo shank skin contain rare tight-junctional strands and only small gap junctions. Exposure of this tissue to retinoic acid (vitamin-A) (20 U/ml) in organ culture, however, induces mucous metaplasia, accompanied by tight-junction formation and gap-junction growth; untreated specimens continue to keratinize. To investigate sequential stages of junctional assembly and growth, we examined thin sections and freeze-fracture replicas at daily intervals for 3 days. During the metaplastic process, tight junctions assemble in midepidermal and upper regions, beginning on day 1 and becoming maximal on day 3. Two tight-junctional patterns could be tentatively identified as contributing to the emergence of fully formed zonulae occludentes: (a) the formation of individual ridges along the margins of gap junctions; (b) de novo generation of continuous ramifying strands by fusion of short strand segments and linear particulate aggregates near cellular apices. Gap junction enlargement, already maximal at day 1, occurs primarily three to four cell layers deep. Growth appears to occur by annexation of islands of 20-40 8.5-nm particles into larger lattices of islands separated by particle-free aisles. Eventually, a single gap junction may occupy much of the exposed membrane face in freeze-fractured tissue, but during apical migration of the cells such junctions disappear. The vitamin- A chick-skin system is presented as a responsive model for the controlled study of junction assembly.

Animals↗

PubMed1991

Neuroblastoma grafts are noninvasively removed within mouse neocortex by selective laser activation of intracellular photolytic chromophore.

Studies of neural cell transplantation would be aided by the ability to damage or destroy, noninvasively and extremely selectively, grafted cells at defined times following their initial implantation. Mechanisms of graft integration and performance could be investigated, also providing insight into natural injury and repair mechanisms. At long wavelengths between 650 and 850 nm, laser energy can penetrate several millimeters of brain tissue without absorption or damage to the unpigmented tissue. However, targeted cells are selectively damaged by illumination at these long wavelengths if they contain latex nanospheres with incorporated cytolytic chromophores (e.g., chlorin e6). Light penetration allows many thousands of cells to be lesioned simultaneously, noninvasively, and deep within a surrounding matrix of other tissue. Such laser-activated damage has been termed laser photolysis (PL). We studied damage to C1300 neuroblastoma (NB) cells grafted into mouse neocortex in vivo by this process of PL. NB cells provided a simple and reproducible model of neural grafting, allowing direct histologic assessment of cellular growth and viability by distinct morphologic and mitotic criteria. Cells were cultured by standard methods, labeled in vitro by brief exposure to nanospheres containing chlorin e6, and grafted to sites within deep layers of mouse neocortex. Mice were exposed to transcranial, fractionated, unfocused pulses of 670-nm-wavelength energy totaling 90-120 J/cm2. We histologically assessed graft growth and cellular viability over a period from 2 d to 4 weeks, measured graft volumes quantitatively during the period of early rapid growth in controls (2 and 7 d), and generated 3-D reconstructions from serial sections to assist in visual analysis.(ABSTRACT TRUNCATED AT 250 WORDS)

Animals↗