本頁只刊出中文翻譯與中文說明；英文原文請見下方原文連結。
原文連結
論文資訊
- 類型：已發表論文
- 日期：2025-04-07
摘要
Motivation Unraveling the human interactome to uncover 疾病-specific patterns and discover drug targets hinges on accurate 蛋白質-蛋白質 interaction (PPI) predictions. However, challenges persist in 機器學習 (ML) models due to a scarcity of quality hard negative samples, shortcut learning, and limited generalizability to novel 蛋白質s.Results In this study, we introduce a novel approach for strategic sampling of 蛋白質-蛋白質 noninteractions (PPNIs) by leveraging higher-order 網絡 characteristics that capture the inherent complementarity-driven mechanisms of PPIs. Next, we introduce Unsupervised Pre-training of Node Attributes tuned for PPI (UPNA-PPI), a high throughput sequence-to-function ML pipeline, integrating unsupervised pre-training in 蛋白質 representation learning with Topological PPNI (TPPNI) samples, ca
※ 此為已發表論文，全文需透過期刊付費取得