HyMM: hybrid method for disease-gene prediction by integrating multiscale module structure

Area of Science:

• Computational biology
• Systems biology
• Genomics

Background:

• Identifying disease-related genes is crucial for understanding complex diseases.
• Biomolecular networks exhibit module structures, perturbations of which are linked to disease.
• Effectively mining and utilizing these module structures for disease gene prediction remains a challenge.

Purpose of the Study:

• To propose a hybrid method, HyMM, for disease gene prediction by integrating multiscale module structures.
• To enhance the prediction of disease-related genes by utilizing both local and global network information.

Main Methods:

• HyMM employs multiscale modularity optimization with exponential sampling to extract module partitions across different scales.
• Gene disease relatedness is estimated based on the abundance of known disease genes within identified modules.
• A probabilistic model integrates predictions from multiscale partitions and network propagation, further refined by a functional information-based parameter estimation strategy.

Main Results:

• Experiments demonstrate the significance of module partitions at various scales for disease gene prediction.
• HyMM shows stable and superior performance compared to eight state-of-the-art methods.
• Parameter estimation based on functional information further improves HyMM's predictive accuracy.

Conclusions:

• HyMM provides an effective framework for integrating multiscale module structures to enhance disease gene prediction.
• The findings offer valuable insights into the role of multiscale module structures in disease gene identification.
• This approach holds promise for advancing computational biology and disease gene studies.