PubMed Health⌕ Search

SEARCH · PubMed Health

Results for “Deep Learning”

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 415 records · Page 23Linked to original sources
PubMed2026

CRISPGen: A deep generative framework for multi-objective CRISPR/Cas9 guide RNA design via Conditional Latent Diffusion and Dual-Critic Reinforcement Learning.

MOTIVATION: The CRISPR-Cas9 system offers transformative potential for precision genome editing, yet its clinical translation remains constrained by the risk of unintended off-target double-strand breaks. While current discriminative models excel at evaluating pre-specified candidate guides, resolving the fundamental antagonism between on-target cleavage efficiency and off-target specificity within a fixed sequence search space remains a major challenge. RESULTS: We present CRISPGen, a unified deep generative framework that reframes sgRNA design as a multi-objective constrained sequence synthesis problem. It integrates (i) DNABERT-2 genomic-language embeddings, (ii) a conditional latent diffusion generator conditioned on a user-specified on-target efficiency target, and (iii) a dual-critic reinforcement-learning (RL) stage that couples a frozen on-target efficiency critic with a cross-attention off-target discriminator (validation Pearson R=0.8157) trained on a unified corpus of experimental off-target events from six detection platforms. Across 1000 generated sgRNAs, CRISPGen reduces the mean off-target discriminator score by 99.7% relative to the pre-RL baseline and, under an exhaustive whole-genome screen of all 302,631,056 NGG PAM sites in GRCh38, yields zero perfect-match and only 55 one-mismatch genomic hits. We further show, transparently, that the internal on-target critic saturates under RL optimization - an instance of Goodhart's Law - and therefore assess on-target viability using an independent external CRISPRon screen (mean 47.10/100). Repeating the RL fine-tuning stage under three random seeds (with the diffusion generator, DNABERT-2 embeddings, and off-target discriminator held fixed) yields a stable operating point across seeds. Full diversity, per-mismatch, and reproducibility statistics are reported in the Results. AVAILABILITY: Source code is available at https://github.com/malekpouri/CRISPGen; the pre-trained checkpoints and the 3,000,000-sequence library are hosted on Hugging Face (https://huggingface.co/malekpouri/CRISPGen-Checkpoints) and archived on Zenodo under DOI 10.5281/zenodo.21428641.

CRISPR-Cas9↗
PubMed2006

Functional imaging of sequence learning in Parkinson's disease.

Sequence learning, a cognitive task linked to cortico-striatal function, is impaired in Parkinson's disease (PD). We chose this task as a behavioral paradigm to study the functional architecture of PD in treated and untreated conditions. In our studies, participants were scanned with H(2)(15)O while performing a kinematically controlled motor sequence learning task and a matching motor baseline task. Experiments revealed that a specific sequence learning network predicts learning in normal subjects, and in independent cohorts of early and advanced PD patients. The analysis of the relationship of network activity to learning performance revealed diverging influences of dopaminergic therapy and deep brain stimulation (DBS). DBS of the internal GP and of STN increased network activity and task performance, while levodopa decreased both measures. In separate studies, we investigated the role of dopaminergic modulation on brain activation during sequence learning. In healthy subjects dopamine transporter (DAT) binding correlated with learning-related brain activation in prefrontal, premotor and cingulate cortices, and in the thalamus. By contrast, in PD most of these regional relationships were lost. Only ventral and dorsolateral prefrontal cortex activation correlated with caudate dopaminergic input. In a final set of studies, we found a significant decline in learning performance in early stage PD patients followed over the course of 2 years. Longitudinal declines in learning-related activation were found in parietal areas, while concomitant increases were localized to the left hippocampus. These observations support hypotheses on disease-stage and task-specific effects within the different cortico-striato-pallido-thalamocortical loops and the mesocortical system in PD.

Aged↗
PubMed2001

The shortened Study Process Questionnaire: an investigation of its structure and longitudinal stability using confirmatory factor analysis.

BACKGROUND: The Study Process Questionnaire (SPQ) is a widely used measure of learning approach and was proposed to have three orientations: surface, deep, and achieving, each with an underlying motive and strategy. AIMS: This study aimed to examine the factor structure and longitudinal stability over five to seven years of a modified shortened 18-item version of the SPQ. SAMPLES: A total of 1349 medical students completed the shortened SPQ at application and in their final year of medical school. Three additional cohorts of students completed the shortened SPQ during their third and fourth year of medical school (sample size: 194, 203, 174). METHOD: Confirmatory factor analysis was used to examine the dimensionality and longitudinal stability of the shortened SPQ. RESULTS: Like the full 42-item version, the shortened SPQ has six subscales and the data are best fit by three second order shared indicator factors (surface, deep and achieving) and a single higher order composite deep-achieving factor. The longitudinal analysis found 26.8%, 26.3%, and 18.7% of the non-attenuated variance of the surface, deep and achieving factor scores in the final year is predicted from the shortened SPQ completed at application to medical school. CONCLUSIONS: The shortened 18-item SPQ has the same six subscales as the full SPQ as well as three second order shared indicator factors (surface, deep, achieving) and one higher order deep-achieving factor similar to that suggested by Biggs (1987). The longitudinal analysis supports this hypothesis and suggests that these learning approaches are partly stable during medical school undergraduate training and partly modifiable under the influence of the educational environment.

Adult↗
PubMed2002

Experience-dependent changes in cerebellar contributions to motor sequence learning.

Studies in experimental animals and humans have stressed the role of the cerebellum in motor skill learning. Yet, the relative importance of the cerebellar cortex and deep nuclei, as well as the nature of the dynamic functional changes occurring between these and other motor-related structures during learning, remains in dispute. Using functional magnetic resonance imaging and a motor sequence learning paradigm in humans, we found evidence of an experience-dependent shift of activation from the cerebellar cortex to the dentate nucleus during early learning, and from a cerebellar-cortical to a striatal-cortical network with extended practice. The results indicate that intrinsic modulation within the cerebellum, in concert with activation of motor-related cortical regions, serves to set up a procedurally acquired sequence of movements that is then maintained elsewhere in the brain.

Adult↗
PubMed2002

The relation of White Matter Hyperintensities to implicit learning in healthy older adults.

OBJECTIVE: This study examined whether MRI evidence of cerebrovascular disease in the form of white matter hyperintensities (WMH) was associated with decreased implicit sequence learning performance in a high-functioning group of normal elderly volunteers. METHOD: One hundred and eight community-dwelling elderly individuals received an MRI and performed an implicit sequence learning task, the serial reaction time (SRT) task. RESULTS: Hyperintensities present in the white matter were associated with a decreased learning effect. This association was found with both deep white matter and periventricular changes. Other factors affecting SRT performance (i.e., baseline reaction time and switch-cost) were not significantly related to the presence of WMH. CONCLUSIONS: The results indicate that in addition to previously identified generalized cognitive deficits, WMH are also associated with a specific decrease in the implicit learning of sequences.

Aged↗
PubMed2005

Learning in a web-based system in medical education.

New learning environments such as distance education and computer-aided instruction promise to bring a change in today's learning environments by adjusting the relationship between the learner, the educational content and the organization of education. In this study, we explored whether students' approaches to learning related to their perception of a particular virtual learning environment. Scales of the ASSIST questionnaire were loaded in a two-principal component solution, surface and deep-strategic. We found statistically significant correlations between the approaches to learning and the student's attitudes towards ICT. Early identification of approaches to learning and attitudes towards ICT may prove to be important in order to provide assistance to aid the transition of students with diverse individual characteristics and to the design of new learning environments.

Attitude to Computers↗
PubMed1996

The Effects of Cooperative Learning on Student Achievement and Motivation in a High School Geometry Class

In this study, the effects of a form of cooperative group instruction (Student Teams Achievement Divisions) on student motivation and achievement in a high school geometry class were examined. Eighty students were randomly assigned to either a control group receiving traditional instruction or one of two treatment groups receiving cooperative learning instruction. Geometry achievement was assessed using scores from the IOWA Test of Basic Skills and teacher-made exams. An 83-item questionnaire was used as a pretest, posttest, and post-posttest assessment of efficacy, intrinsic valuing, goal orientation, and cognitive processing. Students in the cooperative treatment groups exhibited significantly greater gains than the control group in geometry achievement, efficacy, intrinsic valuing of geometry, learning goal orientation, and reported uses of deep processing strategies. The implications for cooperative group structures and motivation theory are discussed.

Journal Article↗
PubMedDate not supplied

[Chronic abdominal pain: a pediatric approach].

To treat chronic abdominal pain in a child is a difficult task either in a therapeutic or diagnostic point of view, by the variety of its several causes, displays and evolution. It is important not only to the clinician lent also to the child's family to try to identify the possible organic and psychic etiologies; in the organic causes there are frequently gastrointestinal and genitourinary reasons, but there are many dysfunctional mechanisms that shall not be forgotten. The author believes is useful, for a better understanding, to apply Levine and Rappaport's model, that indicates four primary intervening forces of the abdominal pain; 1. predisposition, dysfunction or somatic illness; 2. habits and life style; 3. personality and learned behavior's the necessity of a deep comprehension in the relationship between term case, sometimes frustrating.

Abdominal Pain↗
PubMed2025

GE-IA-NAM: gene-environment interaction analysis via imaging-assisted neural additive model.

MOTIVATION: Gene-environment (G-E) interaction analysis is crucial in cancer research, offering insights into how genetic and environmental factors jointly influence cancer outcomes. Most existing G-E interaction methods are regression-based, which may lack flexibility to capture complex data patterns. Recent advances have investigated deep neural network-based G-E models. However, these methods may be more vulnerable to information deficiency due to challenges such as limited sample size and high dimensionality. Apart from genetic and environmental data, pathological images have emerged as a widely accessible and informative resource for cancer modeling, presenting its potential to enhance G-E modeling. RESULTS: We propose the pathological imaging-assisted neural additive model for G-E analysis (GE-IA-NAM). The flexible and interpretable additive network architecture is adopted to account for individualized effects associated with genetic factors, environmental factors, and their interactions. To improve G-E modeling, an assisted-learning strategy is investigated, which adopts a joint analysis to integrate information from pathological images. Simulations and the analysis of lung and skin cancer datasets from The Cancer Genome Atlas demonstrate the competitive performance of the proposed method. AVAILABILITY AND IMPLEMENTATION: Python code implementing the proposed method is available at https://github.com/Mr-maoge/NAM-IA-GE. The data that support the findings in this article are openly available in TCGA (The Cancer Genome Atlas) at https://portal.gdc.cancer.gov/.

Gene-Environment Interaction↗
PubMed2001

The revised two-factor Study Process Questionnaire: R-SPQ-2F.

AIM: To produce a revised two-factor version of the Study Process Questionnaire (R-SPQ-2F) suitable for use by teachers in evaluating the learning approaches of their students. The revised instrument assesses deep and surface approaches only, using fewer items. METHOD: A set of 43 items was drawn up for the initial tests. These were derived from: the original version of the SPQ, modified items from the SPQ, and new items. A process of testing and refinement eventuated in deep and surface motive and strategy scales each with 5 items, 10 items per approach score. The final version was tested using reliability procedures and confirmatory factor analysis. SAMPLE: The sample for the testing and refinement process consisted of 229 students from the health sciences faculty of a university in Hong Kong. A fresh sample of 495 undergraduate students from a variety of departments of the same university was used for the test of the final version. RESULTS: The final version of the questionnaire had acceptable Cronbach alpha values for scale reliability. Confirmatory factor analysis indicated a good fit to the intended two-factor structure. Both deep and surface approach scales had well identified motive and strategy subscales. CONCLUSION: The revision process has resulted in a simple questionnaire which teachers can use to evaluate their own teaching and the learning approaches of their students.

Factor Analysis, Statistical↗
PubMed1996

Mode of foraging competition is related to tutor preference in Zenaida aurita.

This study compared the direction of social learning in 2 populations of Barbados Zenaida doves (Zenaida aurita). One population (St. James) is territorial; it competes aggressively with conspecifics but scramble competes with heterospecifics. The other population (Deep Water Harbour) forages in large homospecific flocks. Field observations were conducted to quantify intraspecific and interspecific patterns of foraging association and aggression. Wild-caught doves from both areas were then tested on novel foraging tasks demonstrated by either a conspecific or a heterospecific tutor. In all experiments, St. James doves learned more readily from the heterospecific tutor (Carib grackle -Quiscalus lugubris-), whereas Deep Water Harbour doves learned more readily from the conspecific tutor. The type of competitive feeding interaction in the field (i.e., scramble vs. interference) appears to better predict the pattern of social learning in an experiment than does species identity.

Aggression↗
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↗
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↗
PubMed2006

A new locus for synaptic plasticity in cerebellar circuits.

Experimental and computational analyses of cerebellar function indicate that excitatory synapses onto deep nucleus neurons are likely to be a critical site of plasticity during motor learning. In this issue of Neuron, Pugh and Raman report that unconventional stimulus protocols can drive synaptic plasticity in the deep cerebellar nuclei.

Animals↗
PubMed2001

Motor skills and motor learning in Lurcher mutant mice during aging.

Motor learning abilities on the rotorod and motor skills (muscular strength, motor coordination, static and dynamic equilibrium) were investigated in three-, nine-, 15- and 21-month-old Lurcher and control mice. Animals were subjected to motor training on the rotorod before being subjected to motor skills tests. The results showed that control mice exhibited decrease of muscular strength and specific equilibrium impairments in static conditions with age, but were still able to learn the motor task on the rotorod even in old age. These results suggest that, in control mice, efficiency of the reactive mechanisms, which are sustained by the lower transcerebellar loop (cerebello-rubro-olivo-cerebellar loop), decreased with age, while the efficiency of the proactive adjustments, which are sustained by the upper transcerebellar loop (cerebello-thalamo-cortico-ponto-cerebellar loop), did not. In spite of their motor deficits, Lurcher mutants were able to learn the motor task at three months, but exhibited severe motor learning deficits as soon as nine months. Such a deficit seems to be associated with dynamic equilibrium impairments, which also appeared at nine months in these mutants. By two months of age, degeneration of the cerebellar cortex and the olivocerebellar pathway in Lurcher mice has disrupted both lower and upper transcerebellar loops. Disruption of the lower loop could well explain precocious static equilibrium deficits. However, in spite of disruption of the upper loop, motor learning and dynamic equilibrium were preserved in young mutant mice, suggesting that either deep cerebellar nuclei and/or other motor structures involved in proactive mechanisms needed to maintain dynamic equilibrium and to learn motor tasks, such as the striatopallidal system, are sufficient. The fact that, in Lurcher mutant mice, motor learning decreased by the age of nine months suggests that the above-mentioned structures are less efficient, likely due to degeneration resulting from precocious and focused neurodegeneration of the cerebellar cortex. From this behavioral approach of motor skills and motor learning during aging in Lurcher mutant mice, we postulated the differential involvement of two transcerebellar systems in equilibrium maintenance and motor learning. Moreover, in these mutants, we showed that motor learning abilities decreased with age, suggesting that the precocious degeneration of the cerebellar Purkinje cells had long-term effects on motor structures which are not primarily affected. Thus, from these results, Lurcher mutant mice therefore appear to be a good model to study the pathological evolution of progressive neurodegeneration in the central nervous system during aging.

Aging↗
PubMed2024

Prediction of metabolic syndrome using machine learning approaches based on genetic and nutritional factors: a 14-year prospective-based cohort study.

INTRODUCTION: Metabolic syndrome is a chronic disease associated with multiple comorbidities. Over the last few years, machine learning techniques have been used to predict metabolic syndrome. However, studies incorporating demographic, clinical, laboratory, dietary, and genetic factors to predict the incidence of metabolic syndrome in Koreans are limited. In the present study, we propose a genome-wide polygenic risk score for the prediction of metabolic syndrome, along with other factors, to improve the prediction accuracy of metabolic syndrome. METHODS: We developed 7 machine learning-based models and used Cox multivariable regression, deep neural network (DNN), support vector machine (SVM), stochastic gradient descent (SGD), random forest (RAF), Na&#xef;ve Bayes (NBA) classifier,&#xa0;and AdaBoost (ADB) to predict the incidence of metabolic syndrome at year 14 using the dataset from the Korean Genome and Epidemiology Study (KoGES) Ansan and Ansung. RESULTS: Of the 5440 patients, 2,120 were considered to have new-onset metabolic syndrome. The AUC values of model, which included sex, age, alcohol intake, energy intake, marital status, education status, income status, smoking status, dried laver intake, and genome-wide polygenic risk score (gPRS)&#xa0;Z-score based on 344,447 SNPs (p-value&#x2009;<&#x2009;1.0), were the highest for RAF (0.994 [95% CI 0.985, 1.000]) and ADB (0.994 [95% CI 0.986, 1.000]). CONCLUSIONS: Incorporating both gPRS and demographic, clinical, laboratory, and seaweed data led to enhanced metabolic syndrome risk prediction by capturing the distinct etiologies of metabolic syndrome development. The RAF- and ADB-based models predicted metabolic syndrome more accurately than the NBA-based model for the Korean population.

Humans↗