graph neural network

GrapHiC improves Hi-C read imputation by integrating Hi-C and ChIP-seq data in a graph framework, enhancing performance on sparse and cross-cell-type inputs.

This study benchmarks graph neural networks for RNA secondary structure prediction, showing they outperform traditional energy-based methods.

Pair-EGRET predicts protein-protein interaction sites using graph attention networks and protein language models, outperforming existing methods on experimental data.

The paper predicts mitosis stages using a CNN followed by a GNN, incorporating timestamps as a graph to model internal interactions, achieving performance comparable to state-of-the-art methods.

BRDKRM classifies drugs as disease-modifying or symptomatic using a graph-based framework, achieving 86.78% accuracy and identifying 530 genes involved in both categories.

MedGraphNet integrates multi-relational biological data with text knowledge to improve biomedical predictions, outperforming single-relation methods and identifying drug side effects without direct training.

GNN4DM identifies overlapping functional disease modules by integrating network topology and genomic data, outperforming existing methods and discovering novel multimorbidity modules.

This paper introduces scSimGCL, a graph contrastive learning framework for scRNA-seq data that improves cell clustering through self-supervised pretraining and high-quality representation learning.

This paper introduces HGGEP, a hypergraph neural network model that predicts gene expression from histology images by capturing cell morphology and higher-order feature associations.

siRNADesign uses a Graph Neural Network to predict siRNA efficacy by analyzing RNA sequences and interactions, achieving state-of-the-art performance in both computational and experimental settings.

CAST enables spatial-to-spatial searching and matching of single-cell spatial omics data across technologies and conditions, identifying disease-associated pathways and translational efficiency variations.

GGAssembler designs and synthesizes cost-effective combinatorial mutation libraries for protein engineering using a graph-theoretical approach.

This paper uses molecular simulations and a graph neural network to identify potential methanogenesis inhibitors and their biochemical properties.

This paper presents a unified computational model that generates various spatial and object-sensitive neurons in the hippocampus by integrating visual and path integration information through a Graph Neural Network.

The paper introduces graph diffusion wavelets to estimate functional connectivity from structural connectivity, achieving high correlation and revealing power-law diffusion scales and frontal pole dominance.

This paper discovers a novel quinoline-based PLpro inhibitor, GZNL-2016, that remains effective against drug-resistant SARS-CoV-2 variants like Omicron BA.5 and XBB.1.

This paper introduces GearBind, a geometric graph neural network for in silico antibody affinity maturation, achieving significant improvements in binding affinity.

This paper introduces MICAH, a graph neural network framework that identifies cancer-associated microbial communities in intratumoral microbiome data across five cancer types.

This paper proposes a graph neural network approach for phenotype prediction that incorporates protein-protein interactions, outperforming existing methods in Alzheimer's disease prediction.

The paper introduces a graph neural network framework that integrates multi-omics data with biological priors to identify biomarkers in Alzheimer's disease.

The paper introduces a graph neural network to predict small-molecule inhibition of protein complexes using compound and protein structure inputs, achieving high accuracy on benchmark datasets.

KGRACDA predicts circRNA-disease associations by capturing local depth information in a knowledge graph, outperforming existing methods.

This paper introduces GSALIDP, a model that predicts dynamic interactions of intrinsically disordered proteins using graph and sequence-based methods, achieving performance comparable to or better than existing models.

MOGDx integrates multi-omic data using a graph convolutional network to classify heterogeneous diseases, achieving state-of-the-art performance and identifying relevant genomic markers.

This paper proposes a deep graph neural network to predict one-dimensional hepatic vascular hemodynamics, achieving high accuracy and speed compared to CFD methods.

Human-in-the-loop labeling of nuclear live-cell movies produced DynamicNuclearNet (over 2,621 divisions); Caliban, a tracker trained on it, beat four published trackers at division detection.

This paper introduces laruGL-PPI, a label-aware hierarchical subgraph learning method for protein-protein interaction prediction that improves performance and efficiency by addressing topological shortcuts and computational costs.

This paper introduces a graph database approach to expand fragment merging by incorporating bioisosteric substructures, improving computational efficiency and chemical diversity in drug design.

This study introduces mosGraphFlow, a graph AI model for analyzing multi-omic data to identify AD-associated biomarkers and signaling pathways, achieving high classification accuracy.

DeepGP predicts glycopeptide MS/MS spectra and retention time using a hybrid deep learning framework, improving glycoproteomics analysis by distinguishing similar glycoforms and increasing detection sensitivity.