HGCJAMH: A Method for circRNA-Drug Sensitivity Prediction Based on Higher-Order Moment-Guided Model and Hypergraph Jumping Learning Mechanism.

Circular RNAs (circRNAs) play a key role in regulating cellular drug sensitivity, and they hold significant promise as potential biomarkers in disease treatment and precision medicine. However, existing circRNA-drug sensitivity association (CDSA) prediction methods generally face the following challenges: high dependence on wet experiments, sparse association data leading to a high percentage of false negatives, insufficient expression of embedded features, and inadequate modeling of higher-order heterogeneous relationships. To solve the above problems, this paper proposes a prediction model, HGCJAMH, based on the higher-order moment-guided model and hypergraph jump learning mechanism, which introduces the KNN and K-means algorithms to construct a multiview hypergraph and models the higher-order complex relationships between circRNAs and drugs. Moreover, it effectively enhances the discriminative power and stability of feature representations through a high moment-guided convolution mechanism and skip-graph contrastive learning. After that, the feature attention mechanism and hierarchical multi-view fusion module are further introduced to realize the adaptive integration and enhanced expression of information from different views. Finally, the association prediction is realized by neural network matrix complementation. In the 5-fold cross-validation, the HGCJAMH model achieves 98.19% AUC and 98.18% AUPR, which is significantly better than several existing mainstream models. The ablation experiments and case validation show that the proposed method not only has superior performance but also has good biological interpretability, demonstrating great potential in the CDSA prediction task. The source code and dataset are available at https://github.com/Forthedark-web/HGCJAMH.