EMGMDA: a multi-modal graph neural framework for robust prediction of miRNA-disease associations.

The results show that EMGMDA achieves statistically significant improvements over all baselines: on HMDD v2.0, the Holm-adjusted p -values are < 0.01 for both AUC and AUPRC across all comparisons; on HMDD v3.2, EMGMDA is highly significant against all baselines on AUC (Holm-adjusted p -values < 0.001) and remains significant on AUPRC (Holm-adjusted p-values < 0.01, with comparisons against MDFormer and DARSFormer reaching Holm-adjusted p -values < 0.001).

also in 132,142 other papers