Dynamic Module Detection in Temporal Attributed Networks of Cancers.

Area of Science:

• Computational biology
• Bioinformatics
• Cancer research

Background:

• Tracking dynamic modules in cancer progression is crucial for understanding pathogenesis, diagnosis, and therapy.
• Existing algorithms for dynamic module detection in temporal cancer networks lack integration with heterogeneous genomic data, leading to suboptimal performance.

Purpose of the Study:

• To propose a novel algorithm, Temporal Attributed Network Matrix Factorization (TANMF), for detecting dynamic modules in cancer temporal attributed networks.
• To integrate temporal network information and gene attributes into a unified objective function for robust dynamic module detection.

Main Methods:

• Developed TANMF, an algorithm that incorporates temporality and gene attributes into an objective function, framing dynamic module detection as an optimization problem.
• TANMF jointly decomposes network snapshots from consecutive time steps, fusing attributes via regulations and applying an L1 constraint for enhanced robustness.
• Applied the algorithm to breast cancer data to identify dynamic modules and assess their biological significance and clinical relevance.

Main Results:

• TANMF demonstrates superior accuracy compared to state-of-the-art methods in detecting dynamic modules.
• Dynamic modules identified by TANMF in breast cancer data are significantly enriched in known biological pathways.
• The detected dynamic modules show a strong association with patient survival time, highlighting their clinical relevance.

Conclusions:

• TANMF offers an effective approach for the integrative analysis of heterogeneous omics data in cancer research.
• The algorithm enhances the accuracy of dynamic module detection, providing valuable insights into cancer progression and patient outcomes.
• This work facilitates a deeper understanding of cancer pathogenesis and supports the development of novel diagnostic and therapeutic strategies.