@techreport{oai:ipsj.ixsq.nii.ac.jp:00220086,
 author = {Mayumi, Kamada and Atsuko, Takagi and Ryosuke, Kojima and Yasushi, Okuno and Mayumi, Kamada and Atsuko, Takagi and Ryosuke, Kojima and Yasushi, Okuno},
 issue = {3},
 month = {Sep},
 note = {Genomic medicine is expected to play a key role in realizing precision medicine. On the other hand, only a few percent of the huge number of genomic variants detected by genome analysis have clear clinical significance. Many computational methods have been developed to predict the pathogenicity of variants. Although the mechanisms of diseases involve complex biomolecular interactions, conventional methods have not been able to consider the molecular relationship. In this study, we developed PathoGN, which represents the molecular network as graphs and predicts the pathogenicity of variants using Graph Convolutional Networks. In performance evaluation using benchmark sets, PathoGN outperformed existing methods. Results of evaluation with ClinVar, a database of disease-related variants, PathoGN was shown to accurately predict the current label of a variant whose significance was unknown in the past. These results suggest the usefulness of considering biological networks., Genomic medicine is expected to play a key role in realizing precision medicine. On the other hand, only a few percent of the huge number of genomic variants detected by genome analysis have clear clinical significance. Many computational methods have been developed to predict the pathogenicity of variants. Although the mechanisms of diseases involve complex biomolecular interactions, conventional methods have not been able to consider the molecular relationship. In this study, we developed PathoGN, which represents the molecular network as graphs and predicts the pathogenicity of variants using Graph Convolutional Networks. In performance evaluation using benchmark sets, PathoGN outperformed existing methods. Results of evaluation with ClinVar, a database of disease-related variants, PathoGN was shown to accurately predict the current label of a variant whose significance was unknown in the past. These results suggest the usefulness of considering biological networks.},
 title = {Network-based pathogenicity prediction for genomic variants},
 year = {2022}
}