Attention-aware Rare Disease Diagnosis via Graph Attention Neural Network

Monogenic genetic diseases, particularly rare ones, pose a significant challenge to human health.Although next-generation sequencing technology has greatly improved the diagnostic rate of these diseases, the process remains time-consuming and labor-intensive.Phenotype-driven diagnostic tools are evolving rapidly and supporting clinical practitioners.In this study, we proposed a diagnostic model based on a graph attention network.This model uses attention mechanisms to uncover potential relationships between genes and phenotypes, and generates a ranked list of candidate genes for identifying causal gene.Our model was tested on three standard datasets and demonstrated an excellent performance.