graph neural network

This study uses graph theory to analyze how temperature-dependent noncovalent interactions in aldolase B and its A149P mutant form a grid-like network that regulates structural thermostability and functional thermoactivity.

This study assesses the combinatorial diversity of aureochrome bZIPs by analyzing dimerization partners and stability using bioinformatics and network science methods.

This paper develops two deep learning models using protein sequence and structure to classify antibiotic resistant genes, achieving over 73% accuracy across 18 categories.

This study introduces a graph-matching based metric to quantify functional connectome differences between individuals and finds it correlates with age, cognitive performance, and familial relationships.

This paper introduces biologically-constrained graph neural networks to decode cognitive processes, revealing functional gradients from sensation to cognition and showing improved inter-subject alignment and behavioral prediction.

DENVIS is a virtual screening method using graph neural networks that outperforms docking-based and sequence-based models by avoiding docking steps and using atomic and surface protein features.

This paper introduces scGCO, a method for identifying spatially variable genes in single-cell data, showing superior performance and scalability.

This study introduces a novel interpretable k-hop graph attention network to integrate multi-omics data and infer essential targets and signaling networks, showing improved performance over existing models.

This paper introduces graph-based algorithms to efficiently compute Laplace transformed coalescence time distributions, improving computational efficiency for population genetic analyses.

Struct2Graph predicts protein-protein interactions using 3D structural data with high accuracy and identifies key residues involved in interactions.

The paper introduces a node embedding algorithm for drug repurposing that outperforms walk-based methods by 53.33% in average precision.

omicsGAT integrates graph-based learning with attention mechanisms to improve cancer subtype analysis using RNA-seq data.

The paper introduces GEMsort, a multichannel neural spike sorting algorithm that reduces data volume and improves sorting speed using signal processing modifications.

The paper uses a deep learning approach to discover novel mosquito repellents by training a graph neural network on historical data and testing candidates in behavioral assays.

This study uses graph neural networks to improve TIL classification on whole slide images by leveraging contextual features from immunofluorescence labeling.

The paper uses c-fos mapping and network analysis to show that alcohol dependence alters brain network activity during abstinence and reaccess in mice, with increased network modularity in high-drinking mice after reaccess.

This paper introduces a graph neural network-based method to predict RNA-protein interactions using CLIP-seq data, showing it can reconstruct interaction networks from minimal information.

This paper introduces a geometric deep learning framework for drug repositioning that leverages heterogeneous information networks to predict drug-disease associations.

This paper introduces a graph convolutional network for analyzing nucleic acid sequences, outperforming existing methods in phage-related tasks.

This paper shows that counting T-cell receptor repertoires outperforms existing methods in detecting CMV and HLA alleles, with an attention model and a simplified graph approach improving accuracy.

This paper introduces a self-supervised learning method, ATMOL, for molecular property prediction using attention-wise graph masking and contrastive learning, achieving state-of-the-art performance.

The paper uses a graph neural network to predict Golgi polarity in migrating epithelial cells, finding it correlates with stress orientation and is affected by local stress field coarsening.

SpaTalk infers cell-cell communication from spatial transcriptomic data using a knowledge graph and graph network, outperforming existing methods on public datasets and revealing insights into tissue communication mechanisms.

The paper analyzes the human structural connectome including the brainstem, finding it to be the most highly connected structure across multiple centrality metrics.

KSTAR predicts patient-specific kinase activities from phosphoproteomic data, improving comparison across labs and identifying tissue-specific profiles and treatment responses.

This paper introduces a hierarchical graph convolutional network for predicting compound-protein interactions using molecular structure and protein sequences, showing improved performance over existing methods.

The paper introduces a phylogenetic profiling method using graph neural networks to reconstruct protein interactions across time, showing improved performance over traditional methods.

The paper analyzes hydrothermal vent species assemblage networks in the Northwest Pacific to identify regional connectivity patterns and their implications for conservation.

Hist2ST predicts RNA-seq expression from histology images using a deep learning model combining Transformer and graph neural networks, outperforming existing methods in gene expression prediction and spatial region identification.

The paper introduces a graph-theoretic algorithm to identify CTCF-mediated chromatin loops that are linked in 3D space using ChIA-PET data, finding non-random linked regions with distinct centrality metrics.