Prioritizing Disease Genes Using Multiple Protein Interaction Networks
Author Information
Author(s): Zhang Wangshu, Sun Fengzhu, Jiang Rui
Primary Institution: Tsinghua University
Hypothesis
Can a Bayesian regression approach effectively integrate multiple protein-protein interaction networks to prioritize disease genes?
Conclusion
The Bayesian regression approach significantly outperforms existing methods in prioritizing disease genes by effectively integrating multiple protein-protein interaction networks.
Supporting Evidence
- The Bayesian regression approach shows a linear relationship between disease similarity and gene proximity.
- The method outperforms the existing CIPHER approach and ordinary linear regression in prioritizing disease genes.
- Integration of multiple PPI networks improves the coverage of disease gene prioritization.
Takeaway
This study shows that by using a smart math method, we can find important genes that cause diseases by looking at how genes are connected to each other.
Methodology
The study uses a Bayesian regression model to analyze gene proximities from multiple protein-protein interaction networks and validate the results through leave-one-out cross-validation.
Potential Biases
The integration approach does not appear to be biased towards well-characterized genes, as the analysis shows no significant differences in ranks based on the number of networks a gene appears in.
Limitations
The method's performance may drop when integrating multiple networks due to missing data, as genes not present in a network are assigned a proximity of zero.
Statistical Information
P-Value
p<2.2E-16
Statistical Significance
p<2.2E-16
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website