graph neural network

xSiGra is a graph-based AI model that identifies interpretable features in spatial single-cell data, showing that cell activity is influenced by neighboring cells and revealing endothelial-tumor interactions.

This paper introduces SSGATE, a dual-path graph attention auto-encoder for integrating single-cell and spatial multi-omics data, showing improved performance and robustness over existing methods.

DeepEnzyme improves kcat prediction using Transformer and GCN with sequence and 3D structure features, showing robustness for low similarity enzymes and mutation impact analysis.

This paper introduces XVGAE, a cross-view graph neural network that integrates gene expression, spatial coordinates, and histological images to identify spatial domains, achieving better clustering accuracy and biologically meaningful results.

Gra-CRC-miRTar is a pre-trained graph neural network framework for identifying miRNA targets in colorectal cancer, achieving better performance than existing methods and revealing functional interactions.

OmicVerse integrates bulk and single-cell RNA-seq analysis with a pipeline that restores continuity in interrupted single-cell data using Beta-Variational AutoEncoder and graph neural networks.

The paper introduces SDAN, a deep learning method that integrates gene annotations with scRNA-seq data to improve cell classification and gene set identification, outperforming existing methods in three applications.

This paper develops a graph neural network method to predict spatio-temporal stochastic gene expression efficiently, achieving faster computation and extrapolation capabilities.

This paper identifies biosignatures specific to different COVID-19 disease phases through integrative multi-omics network analysis and finds that disease severity correlates with increased biosignature interactions across omics data.

The paper introduces a graph neural network for predicting protein-ligand interactions without structural data, achieving competitive performance using representation learning.

RevGraphVAMP improves protein simulation analysis by combining graph neural networks and physical constraints, achieving higher accuracy in dimensionality reduction and state transition prediction.

The paper improves a graph-based method for evaluating cell cluster identities by addressing practical implementation issues and enhancing utility with new features and techniques.

SCREEN is a graph neural network that predicts catalytic residues and assesses mutation tolerance in enzymes by integrating structural and functional information.

The paper uses network analysis to show that in early insulitis, immune cells preferentially interact with β-cells in specific regions of pancreatic islets in NOD mice.

ADMET-AI, a Chemprop-RDKit graph neural network trained on 41 TDC datasets, has the best average leaderboard rank and scores one million molecules in 3.1 hours.

This paper predicts spatial gene expression from histological images using adaptive spatial GNNs, improving performance on breast cancer data.

PEPerMINT imputes missing peptide abundance values in proteomics using graph neural networks, outperforming existing methods across multiple datasets and missingness scenarios.

PSICHIC predicts protein-ligand interactions from sequence data using a physicochemical graph neural network, achieving state-of-the-art accuracy and interpretability in binding affinity and functional effect prediction.

BioPathNet improves link prediction in biomedical knowledge graphs by using path-based reasoning, outperforming existing methods and identifying novel drug indications and gene interactions.

Graspot integrates spatial transcriptomics data using a graph attention network with optimal transport, improving alignment across datasets and revealing tumor microenvironment structures in breast cancer.

This paper identifies optimal drug combinations for tongue neoplasms using a Graph Neural Network within the RAIN protocol, finding Cisplatin, Bleomycin, and Fluorouracil as effective treatments.

This paper introduces TriVIAI, an explainable AI tool for predicting missense mutation pathogenicity using three modalities, achieving high performance metrics and offering biological insights and explainability.

This paper introduces a graph-based model to identify druggable vulnerabilities in solid cancers, finding 43 novel targets by leveraging cooperativity in molecular networks.

This paper introduces AMPGLDA, a method predicting lncRNA-disease associations using adaptive meta-path generation and a deep neural network, achieving high AUC and AUPR scores.

PNAbind predicts protein-nucleic acid binding using graph neural networks, achieving high accuracy in binding site prediction and discrimination between DNA and RNA specificity.

The paper introduces STMask, a graph self-supervised learning method for spatial transcriptomics data, improving spatial clustering and trajectory inference.

PathFinder infers multi-cell signaling pathways and communications using a graph transformer model, identifying novel interactions in AD mouse models.

This paper introduces GRACHIP, a graph neural network model that improves prediction of cell-type-specific 3D genome architectures by incorporating genomic interaction information, showing better accuracy and generalizability than existing methods.

CellWalker2 improves cell type annotation and mapping by integrating multi-omic data and hierarchical relationships, outperforming existing methods in simulations and demonstrating applications in brain and immune system data.

The paper constructs a multimodal canine brain atlas and introduces a hypergraph-based method for fine-grained cortical cytoarchitectonic partitioning.