SEARCH · Search PubMed

Results for “Latent diffusion”

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.

SourceResearch connectionFrom yearThrough year

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

11 recordsLinked 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↗

PubMed2026

Enhancing pan-cancer spatial transcriptomics at single-cell resolution with stPainter.

Subcellular spatial transcriptomics can resolve tissue architecture at cellular scale, but sparse gene panels and limited detection sensitivity constrain downstream analysis. Existing enhancement methods often require tissue-matched single-cell RNA sequencing (scRNA-seq) references and dataset-specific retraining. Here we show that stPainter, a conditional generative model pretrained on a pan-cancer scRNA-seq atlas, can enhance spatial transcriptomics data without matched references or retraining. Using a latent diffusion architecture guided by Stochastic Differential Equations (SDE), stPainter reconstructs expanded expression profiles from sparse measurements and produces latent representations for clustering and cell-state analysis. When we apply stPainter upon 6 spatial transcriptomics datasets of different cancer types, we demonstrate that our model empowers downstream biological analyses, including fine-grained subpopulation clustering and pathway enrichment. Comparison with spatially resolved proteomics (CODEX) provided independent support for regional agreement between imputed cellular compositions and protein-level tissue organization. These results establish stPainter as a scalable approach for analyzing tumor microenvironments without auxiliary sequencing data.

Spatial Transcriptomics↗

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↗

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↗

PubMed2025

Brain Health Loss Mediates the Effect of Infarct Volume on Functional Outcome in Ischemic Stroke.

IMPORTANCE: Brain health may facilitate resilience to detrimental consequences from neurological diseases. Infarct volume is associated with poor functional outcome after acute ischemic stroke (AIS), but potential mediating effects through stroke-related brain health loss have not been investigated. OBJECTIVE: To determine whether stroke-related brain health loss, quantified by change in MRI derived effective Reserve (eR), mediates the effect of acute infarct volume on functional outcome after AIS. DESIGN: Observational multicenter cohort study. SETTING: We analyzed data from the GASROS (n=488) and MRI-GENIE (n=560) cohorts, collected 2003-2011. PARTICIPANTS: Adult patients consecutively diagnosed with AIS, with available admission MRI. EXPOSURE: At admission, white matter hyperintensity (WMH) and normal-appearing brain volumes were assessed on T2-FLAIR, and acute infarct volume on diffusion weighted imaging. WMH was normalized by brain volume, creating WMH load. We quantified brain health using eR, a latent variable incorporating age, WMH load, and normal-appearing brain volume. ΔeR reflected the change in eR when acute infarct volume was included, representing stroke-related brain health decline. Mediation analysis was used to determine if ΔeR mediates the effect of infarct volume on functional outcome (modified Rankin Scale [mRS] at 90 days). MAIN OUTCOME MEASURE: Proportion of mediating effect. RESULTS: We included 1,048 patients (median age 67y, 38% females). At baseline, median NIHSS score was 3 (IQR 1-7), median infarct volume 3.1mL (IQR 0.9-15.5). At 90 days, median mRS score was 1 (IQR 1-3) and 51 (5%) patients had died. In mediation analysis, ΔeR significantly mediated 36% (95% CI 16-56%) of the total effect of infarct volume on functional outcome (direct effect (ß=0.15 [95% CI 0.09-0.22], p<0.001; indirect effect mediated through ΔeR: ß=0.09 [95% CI 0.04 to 0.14], p=0.001). In subgroup-analyses, the mediative effect was apparent among female but not male, and among patients aged >67y but not ≤67y. CONCLUSIONS AND RELEVANCE: Stroke-related structural brain health loss mediates about one third of the effect of acute infarct volume on functional outcome after ischemic stroke, with important sex and age differences. Brain health significantly influences outcome and recovery potential, and may be considered a key biomarker when modeling outcome after AIS.

acute ischemic stroke↗

PubMed1975

[Pulmonary edemas due to acute heroin poisoning].

Their frequency is estimated with difficulty, although on autopsy pulmonary edema is found almost routinely. It is a major complication of overdoses (48 p. 100 of severe intoxications). Their formation can be suspected, when after the first phase of respiratory depressions, with coma, myosis, and a variable latent period, a second attack of respiratory insufficiency occurs with tachypnea, and cyanosis. The chest X-ray shows diffuse alveolar infiltration, sparing the apices. The heart being generally of normal size. Rapid disappearance of this infiltrate (24 to 48 hours) enables the elimination of two diagnoses: pneumonia due to inhalation of gastric fluid, an infectious pneumonia. Their pathogenesis remains very debatable: - in the majority of cases abrupt L.V.F. can be eliminated: -on the other hand it could be an allergic accident of the anaphylactic type, or local liberation of histamine, or a local toxic action on the pulmonary capillaries; - hypoxia, secondary to respiratory depression, could lead to pulmonary edema, by the same mechanism as at altitude; - finally, owing to the central neurological disorders a neurogenic theory can be put forward. Their treatment is essentially a combination of Nalorphine with oxygen therapy (by mask, or if necessary by assisted, controlled ventilation) with prevention of inhalation of gastric fluid (gastric emptying) or curative treatment of possible aspiration by antibiotics, and cortico-steroids. Diuretics can be useful, as well as cardiotonics.

Acute Disease↗

PubMed2026

Phenotyping strategies for chronic overlapping pain conditions and internalizing disorders in Veterans: Prevalence, comorbidity, and latent structure as evidence for construct validity.

Chronic overlapping pain conditions (COPCs), internalizing (INT) disorders, and opioid use disorder (OUD) are common, comorbid, and difficult to phenotype at scale. Electronic health record (EHR) studies commonly define cases using Any Code (AC; ≥1 ICD-9/10 code) and Multiple Code (MC; ≥1 inpatient or ≥2 outpatient codes) phenotyping strategies, but it is unclear whether these thresholds change only case numbers or also the clinical relationships among conditions. This cross-sectional study included approximately 950,000 Million Veteran Program participants with ≥2 visits. AC and MC phenotypes for 17 conditions spanning COPCs, INT, and OUD were compared in prevalence, case characteristics, comorbidity, and latent structure. Random-thinning analysis compared AC-MC differences to case reduction alone. Construct validity was evaluated through correspondence with expected patterns of association and latent organization. Back pain (AC=58.1%; MC=48.1%), major depressive disorder (41.5%; 36.2%), and post-traumatic stress disorder (32.6%; 29.1%) were most prevalent. MC excluded 33.5% of AC cases on average, and MC cases had greater healthcare utilization, diagnostic burden, opioid exposure, and psychiatric medication use than AC-only cases. The observed mean absolute correlation change (mean |Δr|=0.014) was smaller than in all 1000 random-thinning replicates. Both strategies supported a correlated, four-factor model consistent with "Anxious Misery," "Fear," "Diffuse Pain," and "Head Pain" (AC: CFI=0.987, RMSEA=0.015; MC: CFI=0.987, RMSEA=0.014). The MC strategy reduced prevalence and altered case composition but maintained the expected comorbidity and latent organization patterns among conditions. Findings provide evidence of phenotype construct validity and inform selection of EHR phenotyping strategies for epidemiological and genomic research. PERSPECTIVE: Commonly used EHR phenotyping strategies tested in nearly one million Veterans produce broadly similar latent organization across comorbid and prevalent chronic overlapping pain conditions, internalizing disorders, and opioid use disorder. Findings support construct validity and clarify trade-offs involving case inclusion, recorded burden, and healthcare observation in large-scale research.

Chronic overlapping pain conditions↗

PubMed2026

Dissociating behavioral, neural and experiential effects of prefrontal HD-tDCS during conflict resolution.

Inconsistent evidence regarding the cognitive effects of transcranial direct current stimulation (tDCS) highlights the need for more comprehensive approaches to assess its impact. This study aimed to investigate the effects of high-definition tDCS (HD-tDCS) on conflict resolution by combining behavioral, neural, and subjective experience measures. Sixty participants were randomly assigned to anodal, cathodal, or sham HD-tDCS groups and completed a 30-min flanker task. EEG was recorded during the first and last blocks (without stimulation), while stimulation was applied during the intermediate blocks of the task. Using a multidimensional methodological approach including Drift-Diffusion Modeling (DDM), EEG spectral analysis, Lempel-Ziv complexity, and Temporal Experience Tracing (TET), we assessed the cognitive, neural, and phenomenological effects of stimulation. Behavioral results indicated no significant improvements in reaction times or accuracy across the stimulation groups. Similarly, DDM parameters showed no effect of HD-tDCS on latent cognitive processes. However, EEG data revealed a significant reduction in neural complexity in the anodal group during resting-state, suggesting a stabilization or reorganization of neural dynamics. Subjective experience analysis identified two distinct clusters of task-related feelings, though time spent in these experiential states did not differ between groups. Interestingly, sensation of stimulation was significantly higher for anodal stimulation than sham when analyzed as a single dimension. Despite null behavioral effects, this study provides important insights into the neural and subjective responses to HD-tDCS and highlights the value of integrating complementary multidimensional approaches to better characterize brain stimulation effects. These findings contribute to the ongoing debate about the efficacy of tDCS in cognitive enhancement.

Humans↗

PubMedDate not supplied

Mechanisms of lysosomal enzyme release from leukocytes. IV. Interaction of monosodium urate crystals with dogfish and human leukocytes.

In order to determine the possible mechanisms whereby interactions between phagocytic cells and crystals of monosodium urate (MSU) lead to cell death with simultaneous release of both cytoplasmic and lysosomal enzymes, phagocytic leukocytes of the smooth dogfish shart Mustelus canis were studied by means of light and electron microscopy, and biochemistry. Lysosomes of these cells can be stained supravitally with toluidine blue and are large enough (0.7-0.8 mu) to be clearly resolved with the light microscope. Light microscopic observations showed that of cells exposed to MSU 87% of those containing visible ingested crystals died within 1 hour, whereas 92% of adjacent cells in the same wet mount without such srystals survived. Cell death occured after a latent period of 10-15 minutes following fusion of lysosomes with crystal-containing phagosomes. Electron microscopic examination of both dogfish and human leukocytes exposed to MSU for more than 1 hour and then fixed in situ revealed occasional discontinuities or ruptures in secondary lysosome membranes. Endogenous peroxidase activity could be cytochemically localized in primary and secondary lysosomes and in the cytoplasm adjacent to such ruptured secondary lysosomes. It was not seen adjacent to primary lysosomes, a result indicating that the cytoplasmic reaction product was not a diffusion artifact. To exclude the possibility that crystals were exercising their affect primarily upon the plasma membrane, suspensions of dogfish buffy coat cells were incubated with cytochalasin B (5 mug/ml, 10 minutes), which inhibits phagocytosis but not exocytosis of lysosomal enzymes by stimulated phagocytes. Whereas cells exposed to MSU crystals released 30% of their content of lysosomal beta-glucuronidase activity and 28% of their cytoplasmic lactate dehydrogenase (LDH) activity within 3 hours, preincubation with cytochalasin B reduced the release of LDH activity within that period to 6% but reduced the release of beta-glucuronidase activity only to 20%. Preincubation with 10-3 M cyclic adenosine monophosphate (cAMP) and theophylline (10-3 M), which inhibit lysosomal fusion, reduced the release of both LDH and beta-glucuronidase activities to 7% and 6% respectively. Cells that were preincubated with both cytochalasin B and cAMP + theophylline released only 1% LDH activity and 4% beta-blucuronidase activity. These results are compatible with the "suicide sac" hypothesis of lysosomal enzyme release mediated by MSU for the following reasons: a) cell death was seen to follow uptake, not mere exposure to crystals, b) ultrastructural studies indicated that the primary injury was to the secondary lysosome membrane, and c) cell death was reduced when either phagocytosis or lysosomal fusion was inhibited.

Animals↗

PubMed2026

A Knowledge-Enhanced Multimodal Framework with Genomic Reconstruction for DLBCL Drug Response Prediction.

Diffuse large B-cell lymphoma (DLBCL) exhibits substantial biological heterogeneity, leading to pronounced variability in patient response to therapy. Accurate drug response prediction is therefore critical for precision treatment but remains challenging in clinical settings where genomic sequencing, a highly informative modality, is frequently incomplete. Existing methods, often developed from cell-line pharmacogenomic datasets or single-modality data, typically assume fully observed molecular profiles and thus show limited robustness under missing genomic data. To address this limitation, a knowledge-enhanced multimodal framework with genomic reconstruction (KeM-DRP) is proposed for individualized drug response prediction in DLBCL. The framework models the central role of genomics by integrating biological prior knowledge through a gene-pathway-biological process hierarchy, enabling robust representation learning from sparse observations. To compensate for missing genomic measurements, a cross-modal genomic compensation module reconstructs genomically informed latent features from routinely available clinical modalities. Furthermore, a genomics-guided adaptive fusion strategy dynamically integrates heterogeneous modalities conditioned on observed or reconstructed genomic representation. Experiments on a real-world DLBCL cohort demonstrate that KeM-DRP consistently outperforms competitive baselines. The reconstructed genomic representation represents most predictive utility, highlighting the robustness and practical value of the framework under incomplete genomic data.

Journal Article↗

PubMed1975

Chronic antigenic stimulation, herpesvirus infection, and cancer in transplant recipients.

An increased incidence of malignancy has been reported in transplant recipients. The pathogenesis of this increase was originally attributed to immunosuppressive therapy. However, not all tumours are increased in proportion to their occurrence in the general population-75% of reported tumours are lymphorproliferative or carcinoma of the skin, lip, or cervix. This cannot be explained by impaired immunosurveillance, and alternative hypotheses must be considered. 90% of transplant recipients develop clinical or serological evidence of herpesvirus infection. Herpesviruses have been implicated in the pathogenesis of lymphorproliferative tumours and carcinoma of the skin and cervix. They can remain in latent form and be reactivated by allogeneic stimulation and/or immunosuppression. These viruses localise to skin, cervix, and neural tissue-i.e., exactly those sites where cancer develops in transplant patients. Herpesvirus infections in association with the presence of an allogeneic graft in an immunosuppressed patient may be responsible for the increased incidence of both lymphoproliferative tumours and carcinoma of the skin, lip, and cervix in the transplant recipient.

Antibodies, Viral↗