AI Insight
HyperNetWalk is a computational framework that identifies cancer driver genes by working backward from observed gene expression changes to find their upstream causes. The method integrates protein interaction networks with gene regulatory networks in a layered architecture, using reverse random walks to trace dysregulated gene expression back to potential driver mutations. Testing across 12 cancer types from The Cancer Genome Atlas, HyperNetWalk outperformed existing methods at identifying both patient-specific and cohort-level driver genes, successfully detecting both common and rare cancer drivers.
Why it matters
Accurate identification of cancer driver genes is critical for understanding how tumors develop and for selecting targeted therapies for individual patients. This framework's ability to work at both personalized and population levels could improve precision oncology by helping clinicians identify actionable mutations specific to each patient's tumor while also advancing broader understanding of cancer biology.
Understand the Science
Abstract: Accurate identification of cancer driver genes from passenger mutations is essential for understanding tumorigenesis and clinical translation. We present HyperNetWalk, an unsupervised framework that unifies personalized and cohort-level driver gene identification within a shared inference architecture. HyperNetWalk builds a layered signaling-regulatory network by integrating protein-protein interactions, approximating upstream signaling, with a gene regulatory network for downstream transcriptional regulation, with transcription factors serving as interface nodes. Driver identification is formulated as an inverse problem in which observed transcriptional dysregulation is traced back to candidate upstream drivers by reverse random walk. The resulting sample-specific scores are used directly for personalized prediction and as node weights for cross-sample refinement through hypergraph random walk, enabling both local personalized and global cohort-level prediction. Across 12 TCGA cancer types, HyperNetWalk outperformed representative existing methods at both prediction levels. Ablation analyses supported the contributions of the reverse inference formulation and layered network architecture. Further analyses showed that HyperNetWalk captured cancer-type-specific driver signals, prioritized both recurrent and low-frequency candidate drivers, and produced predictions supported by drug-gene interaction and clinical actionability annotations.