graph neural network

The paper identifies how heterogeneous single-cell observations distort manifold-based inference and proposes topological stability descriptors to assess reliability.

This paper introduces a graph-based topic modelling method for single-cell multi-omic data, showing improved performance in clustering and biological interpretation.

GRASP uses a graph neural network to predict RNA subcellular localization by modeling RNA structure, outperforming existing methods across RNA types and transcript lengths.

The paper introduces BOUQUET, an algorithm that identifies 3D-super-enhancers as protein-rich cis-regulatory communities associated with co-activator condensates in mammalian cells.

The paper introduces a dataset and a graph-transformer model to predict IDR-protein binding affinity, achieving high accuracy and generalizing to designed binders.

LinkDTI predicts drug-target interactions using a graph-based framework on a biomedical knowledge graph, outperforming baselines and identifying new interactions.

This study compares two computational models of social perception and inference against fMRI data, finding both explain neural activity in social brain regions.

CAMEX integrates, aligns, and annotates multi-species scRNA-seq data using a GNN, outperforming existing methods and enabling species-specific cell type detection.

The paper introduces ATLAS, a 3D RNA motif library with graph representations and non-Watson-Crick interactions, enabling motif-based RNA structure prediction and design.

This paper presents a multi-modal AI/ML framework using GCNs to predict protein conformations for drug discovery by integrating global and local descriptors.

The paper introduces Mahi, a graph neural network framework that integrates molecular context and functional connectivity to predict gene essentiality and perturbation responses across tissues.

This paper introduces PIGLET, a graph transformer method for drug-target interaction prediction that outperforms existing models on a drug-based split of the Human dataset.

The paper develops a hybrid machine learning framework to identify TDP-43 aggregation inhibitors, discovering two compounds effective in reducing aggregation in cell and model organism assays.

The paper analyzes higher-order repeat (HOR) structures in maize centromeres and knobs, finding they are widespread but not functionally specific to centromeres, while knobs show structured HOR patterns linked to meiotic drive.

CASTLE improves spatial domain detection by integrating cell-type information and contrastive learning, achieving better accuracy across tissues and resolutions.

MutPred-PPI predicts interaction-specific effects of missense variants using graph attention networks, showing strong performance on both seen and unseen proteins.

This study analyzes task-dependent connectivity patterns in honey bees, midges, and birds using graph network modeling, revealing differences in interaction structures across species and behaviors.

BRIDGE improves antimicrobial resistance gene prediction by integrating gene neighbourhood and protein-protein interaction data, achieving 97% classification accuracy.

The paper constructs a cucumber pangenome to analyze structural variant dynamics during domestication and breeding, finding SVs are purged more than SNPs and improving genomic prediction accuracy.

This paper evaluates Graph Neural Networks for brain cortex segmentation using structural MRI and diffusion MRI data, finding GAT models perform competitively with non-graph methods.

The paper uses in silico network perturbation to link exercise to ageing clocks by integrating multi-omic data and experimental validation.

PRISM improves spatial multi-omics analysis by imputing missing data and identifying spatial domains, outperforming existing methods on simulated and real datasets.

SPCoral integrates spatial multi-omics data across different modalities and technologies, improving spatial domain identification and cross-modal analysis.

SpaCEy identifies spatial tissue patterns linked to clinical outcomes using explainable graph neural networks in proteomic cancer data.

This study uses cross-species graph-embedding to show that ageing affects PDAC stroma and therapy response, identifying IRAK4 as a therapeutic target in aged PDAC models.

This study compares the effects of total sleep deprivation and chronic sleep restriction on brain network organization using fMRI and finds they alter functional connectivity differently.

MultiGEOmics integrates multi-omics data using graph-based methods to model cross-omics regulatory signals and handle missing data, showing strong performance across cancer and Alzheimer's datasets.

GNN2Pfam uses graph neural networks to improve Pfam domain annotation by integrating protein sequence and structure data, outperforming HMM-based methods.

The paper develops a graph neural network to predict transcription factors for direct reprogramming by integrating gene regulatory networks with microRNA-mediated regulation.

PAGAN predicts digenic interactions by learning network representations of gene pairs from single-gene data in biological networks.