Mass Enhanced Node Embeddings for Drug Repurposing
Chatzianastasis, M.; Nikolentzos, G.; Vazirgiannis, M.
Show abstract
Graph representation learning has recently emerged as a promising approach to solve pharmacological tasks by modeling biological networks. Among the different tasks, drug repurposing, the task of identifying new uses for approved or investigational drugs, has attracted a lot of attention recently. In this work, we propose a node embedding algorithm for the problem of drug repurposing. The proposed algorithm learns node representations that capture the influence of nodes in the biological network by learning a mass term for each node along with its embedding. We apply the proposed algorithm to a multiscale interactome network and embed its nodes (i. e., proteins, drugs, diseases and biological functions) into a low-dimensional space. We evaluate the generated embeddings in the drug repurposing task. Our experiments show that the proposed approach outperforms the baselines and offers an improvement of 53.33% in average precision over typical walk-based embedding approaches.
Matching journals
The top 2 journals account for 50% of the predicted probability mass.
Similar papers in this journal
- Neural Collective Matrix Factorization for Integrated Analysis of Heterogeneous Biomedical Data 97%
- GraphDTA: Predicting drug-target binding affinity with graph neural networks 96%
- DTI-Voodoo: machine learning over interaction networks and ontology-based background knowledge predicts drug-target interactions 96%
Similar papers in this journal
- Building explainable graph neural network by sparse learning for the drug-protein binding prediction 95%
- Graph-Based Reconstruction and Analysis of Disease Transmission Networks using Viral Genomic Data 93%
- Combined topological data analysis and geometric deep learning reveal niches by the quantification of protein binding pockets 93%
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.