Knowledge Graph Modulated Deep Learning for Limited-Sample Clinical Data Analysis

In three linesGraph-in-Graph (GiG) integrates biological knowledge graphs into deep learning for clinical analysis with limited data. Tested on ~9,700 patients across 5 tasks (cancer detection, prostate diagnosis, pan-cancer classification), GiG outperforms existing methods with gains up to 49 macro-F1 points in limited-sample settings.

Summary generated by Claude — human-verified