PhenoLinker: Phenotype-Gene Link Prediction and Explanation using Heterogeneous Graph Neural Networks
This addresses a critical problem in biology for researchers and clinicians by providing a tool to predict and explain phenotype-gene links, though it appears incremental as it builds on existing graph neural network methods.
The authors tackled the challenge of linking human phenotypes to genetic variants by developing PhenoLinker, a system that uses heterogeneous graph neural networks to score and explain phenotype-gene relationships, aiding in the discovery of new associations and understanding genetic variation.
The association of a given human phenotype to a genetic variant remains a critical challenge for biology. We present a novel system called PhenoLinker capable of associating a score to a phenotype-gene relationship by using heterogeneous information networks and a convolutional neural network-based model for graphs, which can provide an explanation for the predictions. This system can aid in the discovery of new associations and in the understanding of the consequences of human genetic variation.