Advanced Graph Neural Networks

MOSAIC learns graph node embeddings by analyzing dominant spectral components of diffusion operators, outperforming existing methods in biomedical graph tasks.

The paper introduces CLARA, a hierarchical learning framework to mitigate class imbalance in graph data, achieving high F1-score in protein binding site prediction.

This study evaluates how data leakage affects KGE models' performance in biomedical link prediction and finds that current benchmarks may overestimate generalization.

The paper introduces OMNI, a graph neural network framework for biomedical interaction prediction, achieving high accuracy in chemical-gene interaction prediction.

This paper introduces LukePi, a self-supervised pre-training framework for GNNs on biomedical knowledge graphs, which improves link prediction in low-data and distribution-shift scenarios.

The paper introduces GraphEmbed, a scalable network embedding tool that efficiently handles large biological datasets with speed and accuracy improvements over existing methods.

GraFusionNet improves graph classification by integrating node, edge, and semantic features, achieving better performance on toxicity prediction tasks.

This paper introduces a weighted ensemble method for knowledge graph completion that improves drug-disease candidate discovery by combining knowledge graph embeddings and path-based reasoning.

The paper evaluates how knowledge graph topology affects link prediction performance in rare disease variant prioritization, finding that a filtered knowledge graph improves model performance.

PhyGCN improves node representations in hypergraphs using self-supervised learning and variable hyperedge sizes, showing effectiveness on chromatin interactions and polypharmacy side-effects.

The paper proposes an ensemble learning approach to improve link prediction in large biomedical knowledge graphs by sampling subgraphs and aggregating results.

The paper introduces a degree-aware edge sampling method to reduce performance inflation in biomedical graph representation learning caused by imbalanced edge degree distributions.

This paper introduces gGN, a method that represents graph nodes as low-rank Gaussian distributions, showing improved performance in preserving graph structures and measuring semantic similarity.

CLEP generates patient representations by integrating prior biological knowledge with patient data, improving classification performance compared to using raw transcriptomics data.

PecanPy provides a fast and efficient Python implementation of node2vec for embedding nodes in biological networks, improving scalability for dense graphs.

The paper introduces GCN4R, an R library for fitting graph neural networks on networks to generate embeddings for network-based tasks like community detection.

selfRL improves link prediction in heterogeneous biomedical networks by using self-supervised learning, outperforming 25 methods and identifying drugs that inhibit cytokine storms in COVID-19.