Search PubMed⌕ Search

SEARCH · Search PubMed

Results for “Deep Neural Network”

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 37 records · Page 2Linked to original sources
PubMed2025

G4STAB: a multi-input deep learning model to predict G-quadruplex thermodynamic stability based on sequence and salt concentration.

MOTIVATION: G-quadruplexes (G4s) are non-canonical nucleic acid structures formed in guanine-rich regions that modulate gene regulation and genomic stability. The thermodynamic stability of G4s directly influences their biological functions and potential as therapeutic targets. However, current quantitative frameworks for predicting G4 stability rely on predetermined structural features, limiting their effectiveness for diverse G4 topologies, and fail to account for environmental factors such as ion concentration and pH that significantly modulate G4 stability in cellular contexts. RESULTS: We present G4STAB, a multi-input deep learning neural network that accurately predicts DNA G4 melting temperatures based on sequence features, salt concentration, and pH. Trained on 2382 diverse DNA G4 sequences, our model achieves high accuracy (R 2=0.8) without relying on predetermined G4 structural features. G4STAB successfully captures established G4 stability determinants and proposes previously unobserved sequence-stability relationships. Analysis of 391 502 experimentally validated G4s reveals that cancer-like ionic environments alter G4 stability profiles, with a 13.5-fold increase in the number of structures exhibiting physiological melting temperatures (36-42°C). These findings suggest systematic genomic patterns in G4 stability responses across chromosomes and gene types. AVAILABILITY AND IMPLEMENTATION: G4STAB is available at https://github.com/donn-liew/G4STAB; G4STAB web database interface is available at https://donn-liew.github.io/g4stab-web-database/.

G-Quadruplexes↗
PubMed2026

Deep learning reveals genomic regions introgressed between two recurrently hybridizing lynx species.

Recently, diverged species with overlapping distributional ranges have high chances of hybridizing and if hybrids are viable, genomic material can be transferred between species in a process called introgression. To characterize the patterns and consequences of introgression in species with historically low population sizes and recent steep declines resulting in genetic erosion, we analyze the Iberian and Eurasian lynx (EL) as an illustrative and relevant case study. While genome-wide introgression was already detected, here we apply a method using a deep convolutional neural network to detect specific regions of the genome with signals of introgression in three populations of these two species. Over 6% of the genome of both Iberian lynx and ELw shows introgression from the other species, compared with only 2% in the ELs. This observation, along with the results from demographic modeling, suggests that the ELw population is genetically closest to the source of EL introgression, a probably now extinct group that coexisted with the Iberian lynx in Southern Europe and Northern Iberia until recently. As predicted by theory, introgression was generally higher in populations with smaller effective sizes and in genomic regions of high recombination. However, the Iberian lynx did not show higher overall introgression than the more abundant ELw, and coding regions introgressed as frequently as intergenic regions. Local genetic diversity is boosted approximately 3-fold in genomic windows where introgression occurs, potentially including the adaptively relevant and highly diverse MHC region of the Iberian lynx.

Animals↗
PubMed2003

Cortical excitability changes associated with musical tasks: a transcranial magnetic stimulation study in humans.

Neuroimaging studies have suggested differences in cortical activation in human vocalization and musical tasks. However, functional neurophysiological evidence on cortical excitability changes is lacking. We utilized transcranial magnetic stimulation to demonstrate changes in cortical excitability during overt humming and singing tasks. The findings complement those from neuroimaging and support the existence of separate bilateral deep-seated neural networks, as distinct from those for vocalization.

Adult↗
PubMed2025

Development and validation of a deep learning model based on cascade mask regional convolutional neural network to noninvasively and accurately identify human round spermatids.

INTRODUCTION: The difficulty of identifying human round spermatids (hRSs) has impeded applications of the human round spermatid injection (ROSI) technique. RSs can be accurately screened through flow cytometric analysis utilizing the Hoechst fluorescence profile reflecting DNA, but this method is not suitable for isolating hRSs due to the toxicity associated with Hoechst staining. OBJECTIVE: To evaluate the capacity of a deep learning model grounded in a cascade mask region-based convolutional neural network (R-CNN) for the noninvasive and accurate identification of hRSs. METHODS: In this study, we presented the development and validation of a deep learning model for identifying hRSs through the analysis of 3457 optical light microscope images of sorted hRSs obtained via flow cytometric analysis. The model's accuracy and specificity were evaluated by calculating the mean average precision (mAP). Furthermore, a double-blind experiment was conducted to access the reliability of the proposed model in accurately identifying hRSs. It detected the expression of protamine (PRM1) and/or peanut lectin (PNA), which are established markers for RSs. RESULTS: Our deep learning-based model demonstrated a high precision, achieving a mAP of over 0.80 for isolating hRSs in test datasets. The expression of PRM1 and/or PNA was observed in all cells noninvasively selected by our AI model during an independent double-blind test. This phenomenon confirmed the accuracy and effectiveness of the proposed model. The model's capability for noninvasive and accurate isolation of hRSs among spermatogenic cells highlighted its robustness and generalizability for clinical applications. CONCLUSION: The deep learning AI model based on a cascade R-CNN has the ability to accurately identify hRSs among spermatogenic cells. The application of this noninvasive method, which requires no additional procedures in clinical practice, is able to facilitate the widespread implementation of ROSI technique. Therefore, it can provide patients with spermatogenic arrest the opportunity to become biological fathers.

Humans↗
PubMed2000

Micturition and defensive behaviors are controlled by distinct neural networks within the dorsal periaqueductal gray and deep gray layer of the superior colliculus of the rat.

Electrical stimulation of the dorsal periaqueductal gray (DPAG) or the deep gray layer of the superior colliculus (DGSC) of rats placed in an open-field elicited either a display of tense immobility, accompanied by exophthalmus and/or defecation and micturition, or running and jumping responses. Threshold curves of each response were obtained for each structure by the logistic fitting of accumulated response frequencies. DPAG and DGSC threshold curves were compared by likelihood-ratio coincidence tests. The output of micturition was significantly higher following the stimulation of DPAG (P < 0.0005). In contrast, no differences were found for the remaining responses. These data support previous studies in anaesthetized cats suggesting the critical involvement of DPAG in the control of micturition. Furthermore, they also suggest that topographically distinct neural networks within the DPAG and DGSC control micturition and the other defensive behaviors.

Animals↗
PubMed2026

MegaPlantTF: a machine learning framework for comprehensive identification and classification of plant transcription factors.

MOTIVATION: Understanding the role of transcription factors (TFs) in plants is essential for the study of gene regulation and various biological processes. However, both TF detection and classification remain challenging due to the great diversity and complexity of these proteins. Conventional approaches, such as BLAST, often suffer from high computational complexity and limited performance on less common TF families. RESULTS: We introduce MegaPlantTF, the first comprehensive machine learning and deep learning framework for the prediction (TF versus non-TF) and classification (family-level) of plant TFs. Our method employs k-mer-based protein representations and a two-stage architecture combining a deep feed-forward neural network with a stacking ensemble classifier. To ensure robust performance assessment, we report micro-, macro-, and weighted-average performance metrics, providing a holistic evaluation of both frequent and underrepresented TF families. Additionally, we employ threshold-based evaluation to calibrate confidence in TF detection. The results show that MegaPlantTF achieves strong accuracy and precision, particularly with a k-mer size of 3 and a classification threshold of 0.5, and maintains stable performance even under stringent thresholds. In addition to the standard cross-validation tests, a use case study on Sorghum bicolor confirms that our method performs strongly in the genome-wide analysis, making it highly suitable for large-scale TF identification and classification tasks. MegaPlantTF represents a novel contribution by integrating k-mer encoding, binary family-specific classifiers, and a two-stage stacking ensemble into a unified, reproducible framework for large-scale plant TF identification and classification. AVAILABILITY AND IMPLEMENTATION: MegaPlantTF is freely accessible through a public web server available at https://bioinformatics.um6p.ma/MegaPlantTF. The complete source code, including pretrained models and example datasets, is available at https://github.com/Bioinformatics-UM6P/MegaPlantTF.

Transcription Factors↗
PubMed2005

Effects of microinjections of neurotoxin AvTx8, isolated from the social wasp Agelaia vicina (Hymenoptera, Vespidae) venom, on GABAergic nigrotectal pathways.

Several investigations have provided information that defensive behaviors evoked by stimulation of deep layers of the superior colliculus (dlSC) are subjected to inhibitory nigral modulation. This inhibition is made mainly through GABAergic neurons from substantia nigra, pars reticulata (SNpr), that sends outputs toward neural networks of the deep layers of the superior colliculus and dorsal periaqueductal gray matter involved with the organization of fear-like responses. In this work, we compared the effects of two GABAergic agonists, muscimol and baclofen, with the effect of neurotoxin AvTx8 (1567 Da), isolated from the venom of the social wasp Agelaia vicina, microinjected into SNpr of Rattus norvegicus (Wistar rats) prior to dlSC saline or bicuculline microinjections, considering that wasp venom has some influence on the uptake of GABA and/or glutamate neurotransmitters. GABA(A) receptor blockade in the dlSC evoked a vigorous escape behavior, expressed by rapid running, jumps and turns, as compared to control. These defensive reactions were maximized after the intranigral GABA(A) agonism with muscimol, but not after in situ GABA(B) agonism. Nigral microinjection of AvTx8 induced similar effects to those of baclofen, decreasing the intensity of behavioral defensive reactions caused by GABA(A) receptor blockade in the dorsal mesencephalon. These findings suggest that AvTx8 has some effects on GABAergic neurotransmission, increasing the activity of the inhibitory nigro-collicular pathways, causing an anti-panic (antiaversive) effect. Therefore, our work suggests AvTx8 as a novel pharmacological tool to study differences between the two types of GABAergic receptors and excitatory amino acid-mediated mechanisms in the brain and brainstem networks.

Animals↗
PubMed2026

Predictive design of tissue-specific mammalian enhancers that function in the mouse embryo.

Enhancers control tissue-specific gene expression across animals1. Although deep learning2,3 has enabled enhancer prediction and design in mammalian cell lines and non-mammalian model organisms4-10 (reviewed in a previous publication11), it remains unclear whether such approaches can operate within the regulatory complexity of mammalian genomes and tissues in vivo. Here we present a general strategy for designing tissue-specific enhancers that function reliably in mice. We use deep learning to train compact convolutional neural networks on curated chromatin accessibility data and fine-tune them by transfer learning on validated human and mouse enhancers. Guided by these models, we design 15 synthetic enhancers for the heart, limb and central nervous system in mouse embryos, all of which are active in their intended target tissue. These results demonstrate that mammalian enhancer function can be reliably inferred from DNA sequence alone, enabling the predictive de novo design of tissue-specific synthetic enhancers from modest training sets. This work establishes a generalizable framework for programmable control of mammalian gene expression in vivo, opening new avenues in functional genomics, synthetic biology and gene therapy.

Animals↗
PubMed2026

Whole-Genome Deep Learning Predicts Chemotherapy Response in Colorectal Cancer.

Chemotherapy response in colorectal cancer (CRC) exhibits significant heterogeneity, with current clinical predictors failing to capture complex genomic determinants of resistance. We developed a hybrid deep learning framework integrating convolutional neural networks (CNNs) and bidirectional long short-term memory (BiLSTM) networks to analyze whole-genome somatic mutations, evolutionary conservation, chromatin accessibility, and 3D genome architecture in 2,546 TCGA patients. An attention mechanism identified predictive genomic regions. The model achieved an AUC of 0.92 (95% CI: 0.89-0.94) in cross-validation and 0.88 (95% CI: 0.85-0.91) in independent validation, outperforming clinical models (&#x394;AUC = +0.18, p < 0.001). Key predictors included non-coding variants in TP53, KRAS, and PIK3CA regulatory regions. Triple-positive patients (mutations in all 3 regions) had significantly worse progression-free survival (HR = 4.7, p < 0.001). Our framework enables accurate chemotherapy response prediction and reveals novel non-coding resistance mechanisms, advancing precision oncology in CRC.

Humans↗
PubMed2026

GraphyloVar: predicting the impact of non-coding variants using a multi-species sequence model.

MOTIVATION: Understanding the functional impact of genetic variants is a key problem for precision medicine. Tools like CADD, PhyloP, and PhastCons are useful, but they often look at each position in the genome in isolation. This means they can miss important information from the evolutionary history that connects different species. In this paper, we extend our previous model, Graphylo, to predict the effects of variants. Our new model, GraphyloVar, is built to directly utilize the phylogenetic tree that relates the species. RESULTS: GraphyloVar is a deep learning model that considers both DNA sequence and evolutionary patterns from many species. It uses two main components: Graph Convolutional Networks (GCNs) to process the phylogenetic tree, and Transformer encoders to extract features from the DNA sequences. Pre-trained to predict population-level allele frequencies on the TOPMed whole-genome sequencing cohort, GraphyloVar achieves an AUROC of 0.6246 zero-shot on &#x223c;149M held-out variants, and an ensemble with CADD reaches 0.6442 (+0.020, P<10-15). Fine-tuned GraphyloVar achieves the highest AUROC across all 13 MPRA benchmark datasets. By integrating deep learning with explicit phylogenetic input, GraphyloVar offers a powerful and complementary approach to variant effect prediction that utilizes the full evolutionary history from many species to better identify and prioritize important non-coding variants. AVAILABILITY AND IMPLEMENTATION: Code and datasets are available at https://github.com/DongjoonLim/GraphyloVar under DOI: 10.5281/zenodo.20616818.

Phylogeny↗
PubMed2026

SCMO: a deep learning model integrating the single-cell resolution TME ecosystem and multi-omics for survival prediction in CRC patients.

BACKGROUND: Colorectal cancer (CRC) remains a leading cause of global cancer mortality, highlighting the need for precise survival prediction to guide clinical decisions. Although tissue-level multi-omics is widely utilized for survival prediction, its limited resolution cannot capture tumor heterogeneity. Single-cell RNA sequencing (scRNA-seq) enables dissection of the tumor microenvironment (TME) at cellular resolution, supporting personalized prognostic assessment. METHODS: We collected 213 CRC scRNA-seq samples and established a CRC-specific TME atlas comprising 339,060 cells. Using this atlas as a reference, we deconvolved bulk RNA-seq data from TCGA-CRC cohort with the EcoTyper algorithm to reconstruct TME features. Clinical, genomic, and transcriptomic data were obtained from the Xena platform; microbial data were sourced from the BIC database. We integrated TME and multi-omics features through a self-normalizing neural network to construct a deep learning model (single-cell resolution TME ecosystem with multi-omics data [SCMO]) for survival prediction. To enhance interpretability, we utilized the Integrated Gradients algorithm and spatial transcriptomic data to analyze multi-omics and TME features. We performed anticancer drug screening with tumor necrosis factor receptor-associated protein 1 (TRAP1), a critical feature according to the Integrated Gradients algorithm, as a potential target. RESULTS: We identified 13 survival-related TME features from the CRC-specific atlas: 12 cell states and one multi-cellular ecosystem. SCMO, which combined TME and multi-omics features, improved survival prediction and outperformed existing methods, achieving a concordance index of 0.762. The SCMO demonstrated robust performance for long-term predictions, achieving areas under the curve (AUCs) of 0.752, 0.772, and 0.869 for 1-, 3-, and 5-year predictions in the training set, with corresponding test set AUCs of 0.639, 0.756, and 0.772. TME features from the SCMO model revealed that ecosystem density increased with CRC malignancy. Multi-omics features included TRAP1 as a potential drug target. Drug screening identified saikosaponin A as a novel TRAP1 inhibitor, and its anticancer activity was validated in vitro. We developed SCMO-Lite, a simplified model incorporating 12 high-attribution-weight multi-omics features, which demonstrated robust risk stratification. CONCLUSIONS: SCMO combines analytical precision with biological interpretability, offering novel insights for oncology survival prediction.

Humans↗
PubMed1999

Changes in motor network expression related to moulting behaviour in lobster: role of moult-induced deep hypoxia

The well known rhythmically active pyloric neural network in intact and freely behaving lobsters Homarus gammarus was monitored prior to and following ecdysis. Despite long-lasting hormonal and metabolic alterations associated with this process, spontaneous pyloric network activity remained largely unaltered until the last 12-48 h before exuviation. At this time, the most notable change was a progressive lengthening of pyloric cycle period, which eventually attained 500-600 % of control values. It was only in the very last minutes before ecdysis that burst patterning became irregular and the otherwise strictly alternating motor sequence broke down. After the moult, coordinated rhythmicity was re-established within 10 min. Concomitant with these final changes in motor network expression at ecdysis was a drastic reduction in blood oxygen levels which led to a temporary near-anoxia. By imposing similarly deep hypoxic conditions both on intermoult animals and on the pyloric network in vitro, we mimicked to a large extent the moult-induced changes in pyloric network performance. Our data suggest that, despite major surrounding physiological perturbations, the pyloric network in vivo retains stable pattern-generating properties throughout much of the moulting process. Moreover, some of the most significant modifications in motor expression just prior to ecdysis can be related to a substantial reduction in oxygen levels in the blood.

Journal Article↗
PubMed2026

Automated Classification of Lymphoma Subtypes From Histopathological Images Using a U-Net Deep Learning Model: Comparative Evaluation Study.

BACKGROUND: Accurate classification and grading of lymphoma subtypes are essential for treatment planning. Traditional diagnostic methods face challenges of subjectivity and inefficiency, highlighting the need for automated solutions based on deep learning techniques. OBJECTIVE: This study aimed to investigate the application of deep learning technology, specifically the U-Net model, in classifying and grading lymphoma subtypes to enhance diagnostic precision and efficiency. METHODS: In this study, the U-Net model was used as the primary tool for image segmentation integrated with attention mechanisms and residual networks for feature extraction and classification. A total of 620 high-quality histopathological images representing 3 major lymphoma subtypes were collected from The Cancer Genome Atlas and the Cancer Imaging Archive. All images underwent standardized preprocessing, including Gaussian filtering for noise reduction, histogram equalization, and normalization. Data augmentation techniques such as rotation, flipping, and scaling were applied to improve the model's generalization capability. The dataset was divided into training (70%), validation (15%), and test (15%) subsets. Five-fold cross-validation was used to assess model robustness. Performance was benchmarked against mainstream convolutional neural network architectures, including fully convolutional network, SegNet, and DeepLabv3+. RESULTS: The U-Net model achieved high segmentation accuracy, effectively delineating lesion regions and improving the quality of input for classification and grading. The incorporation of attention mechanisms further improved the model's ability to extract key features, whereas the residual structure of the residual network enhanced classification accuracy for complex images. In the test set (N=1250), the proposed fusion model achieved an accuracy of 92% (1150/1250), a sensitivity of 91.04% (1138/1250), a specificity of 89.04% (1113/1250), and an F1-score of 90% (1125/1250) for the classification of the 3 lymphoma subtypes, with an area under the receiver operating characteristic curve of 0.95 (95% CI 0.93-0.97). The high sensitivity and specificity of the model indicate strong clinical applicability, particularly as an assistive diagnostic tool. CONCLUSIONS: Deep learning techniques based on the U-Net architecture offer considerable advantages in the automated classification and grading of lymphoma subtypes. The proposed model significantly improved diagnostic accuracy and accelerated pathological evaluation, providing efficient and precise support for clinical decision-making. Future work may focus on enhancing model robustness through integration with advanced algorithms and validating performance across multicenter clinical datasets. The model also holds promise for deployment in digital pathology platforms and artificial intelligence-assisted diagnostic workflows, improving screening efficiency and promoting consistency in pathological classification.

Humans↗
PubMed2001

Longitudinal changes of metabolites in frontal lobes after hemorrhagic stroke of basal ganglia: a proton magnetic resonance spectroscopy study.

BACKGROUND AND PURPOSE: We investigated serial metabolic changes in frontal lobes of patients with deep intracerebral hemorrhage (ICH) to examine the correlation between N-acetylaspartate (NAA) and degree of motor impairment or clinical outcome. METHODS: - Twenty patients with deep ICH were examined with proton magnetic resonance spectroscopy with the application of a multivoxel method (1 voxel=10x10x20 mm; 64 voxels). NAA/creatine ratios in the white matter of the primary motor and premotor areas on both sides were measured sequentially: within 48 hours, at 2 weeks, and 1 month after onset. The National Institutes of Health Stroke Scale and Barthel Index for disability were measured for each patient. RESULTS: - In the primary motor area on the affected side, where the hematoma did not extend, the NAA/creatine ratio decreased sequentially. At 48 hours and 2 weeks after onset, a negative correlation was detected between NAA/creatine and hematoma volume, but there was no correlation 1 month later. At 2 weeks, NAA/creatine correlated negatively with motor impairment (r=-0.750), and there was a significant correlation with clinical outcome as early as 2 weeks after onset (r=0.954). These sequential changes of NAA/creatine varied according to patients' long-term clinical outcome. Patients with poor outcome demonstrated notable reduction of NAA/creatine over the bilateral frontal lobes. CONCLUSIONS: - The delayed gradual reduction of NAA/creatine ratio in the frontal lobes correlates with motor deficit and clinical outcome after deep ICH, suggesting that the neural networks in the frontal lobe could be important for recovery.

Aged↗
PubMed2026

Artificial Intelligence for Natural Products Discovery and Development.

Natural products (NPs) remain a cornerstone of modern drug discovery, offering stereochemical complexity and diverse bioactivities that precisely modulate therapeutic targets, refined through billions of years of evolution. However, their research has long been hindered by inefficient, empirical workflows, high resource consumption, structural complexity, and the "multicomponent, multi-target" nature of their mechanisms. The exponential growth of genomic, metabolomic, and spectral data has overwhelmed conventional analytical methods, exposing critical bottlenecks in handling high-dimensional, heterogeneous datasets that exceed human interpretive capacity. Artificial intelligence (AI) is emerging as a transformative paradigm to address these challenges, integrating multi-omics and chemical data to shift NP research from fragmented empiricism toward mechanism-driven, precision-oriented development. By leveraging deep learning architectures- including graph neural networks, Transformers, and diffusion-based generative models-AI enables systematic decoding of NP biosynthesis, automated structure elucidation, rational target identification, knowledge extraction from vast unstructured scientific literature, and de novo molecular design. This review comprehensively surveys recent advances in AI applications across the full NP discovery and development pipeline, encompassing genome mining, structure-based and ligand-based virtual screening, multimodal structural characterization, lead optimization, and biosynthetic pathway engineering. We further examine the emerging roles of protein-centric, molecule- centric, and multimodal foundation models, as well as large language models, in bridging genotype-to-chemotype gaps and unlocking unstructured scientific knowledge. Finally, we discuss critical challenges including data scarcity, representational limitations for complex stereochemistry, physical plausibility in generative models, and the urgent need for experimental validation, while outlining future directions toward autonomous experimentation, closed-loop optimization, and human-AI collaborative discovery.

Artificial intelligence↗
PubMed2026

Scalable, generalizable and uncertainty-aware integration of spatial multiomics across diverse modalities and platforms with SCIGMA.

Recent advances in spatial omics technologies have enabled simultaneous profiling of transcriptomic, proteomic, epigenomic, metabolomic and imaging data at high spatial resolution, offering unprecedented opportunities to dissect tissue complexity. However, integrating these diverse and large-scale spatial multimodal datasets remains a major computational challenge. We present SCIGMA, a scalable and generalizable deep learning framework for spatial multiomics integration. SCIGMA introduces an uncertainty-aware contrastive learning objective and multiview graph neural networks to preserve modality-specific signals while learning biologically meaningful joint representations. Unlike previous methods, SCIGMA provides spatially resolved uncertainty estimates, interpretably identifying regions of biological or technical heterogeneity. SCIGMA supports integration of up to five modalities, and its modular framework is extensible to future technologies with even more modalities. It also scales to more than 1 million spatial locations, enabling analysis of high-resolution datasets such as Visium HD and Xenium Prime. We evaluated SCIGMA across 19 datasets spanning 8 modalities, 10 tissues and 9 platforms. On benchmarkable datasets, SCIGMA outperformed other methods in spatial domain detection, modality preservation, feature reconstruction and reproducibility. SCIGMA identifies biologically meaningful structures, refined spatial domains and modality-specific regulatory programs, providing a robust, flexible and future-ready solution for scalable spatial multimodal integration.

Multiomics↗
PubMed2025

CASTER-DTA: Equivariant Graph Neural Networks for Predicting Drug-Target Affinity.

Accurately determining the binding affinity of a ligand with a protein is important for drug design, development, and screening. With the advent of accessible protein structure prediction methods such as AlphaFold, predicted protein 3D structures are readily available; however, methods for predicting binding affinity currently do not take full advantage of 3D protein information. Here, we present CASTER-DTA (Cross-Attention with Structural Target Equivariant Representations for Drug-Target Affinity), which uses an equivariant graph neural network to learn more robust protein representations alongside a standard graph neural network to learn molecular representations to predict drug-target affinity. We augment these representations by incorporating an attention-based mechanism between protein residues and drug atoms to improve interpretability. We show that CASTER-DTA represents a state-of-the-art improvement on multiple benchmarks for predicting drug-target affinity and that it generates novel insights for several related tasks. We then apply CASTER-DTA to create a large resource of the binding affinities of every FDA-approved drug against every protein in the human proteome and make these predictions freely available for download. We also make available a web server for researchers to apply a pretrained CASTER-DTA model for predicting binding affinities between arbitrary proteins and drugs.

deep learning↗
PubMed2026

Integrating histology and spatial transcriptomics via multimodal transformers and contrastive representation learning for accurate gene expression prediction.

Predicting spatial gene expression from Histological images is a fundamental task in understanding tissue organization and molecular phenotypes. However, existing methods often rely on single-model representations or lack effective alignment between image and transcriptomic features. To address these limitations, we propose a unified multimodal learning framework that integrates histological imaging and spatial transcriptomics through a shared latent representation space. Specifically, histological H&E images are encoded by a ResNet50-based convolutional stem and a MobileViT Transformer backbone to extract hierarchical visual representations. Both modalities are projected into a shared latent space via linear-GELU-dropout transformation blocks, enabling cross-modal alignment through a contrastive learning objective that maximizes agreement between the corresponding image and the spot embeddings. Experimental results on the 10x Genomics Visium dataset of human liver tissue demonstrate that MViTGene achieves significantly higher prediction accuracy than existing methods across multiple gene subsets, with improvements of 20%, 33%, and 12% in predicting marker genes, highly expressed genes, and highly variable genes, respectively. The significant improvement in relevance indicates that the model can more accurately capture the true correspondence between tissue morphology and gene expression, therefore enabling more reliable biological interpretation. It provides a computational tool for high-throughput spatial gene expression prediction that balances performance and interpretability.

Humans↗