Graph theory-based measures as predictors of gene morbidity

Area of Science:

• Bioinformatics
• Systems Biology
• Genetics

Background:

• Previous studies suggested a link between protein interaction network properties and gene morbidity.
• A potential bias due to varying amounts of available information for different genes was not accounted for.

Purpose of the Study:

• To model the relationship between gene morbidity and protein interaction network node degrees, controlling for available literature information.
• To investigate the influence of information bias on the observed correlation.

Main Methods:

• Utilized data from Online Mendelian Inheritance in Man (OMIM) and Human Proteome Resource Database (HPRD).
• Quantified gene information availability through PubMed mining.
• Analyzed 7461 genes and 3665 disease identifiers against 9630 nodes and 38756 interactions.

Main Results:

• The correlation between node degree and gene morbidity was found to be significantly influenced by the amount of available information for the gene.
• A positive correlation between node degree and morbidity was observed even after controlling for information factors.

Conclusions:

• The apparent link between protein interaction network structure and gene morbidity may be substantially explained by literature information bias.
• While a residual correlation might exist, it should be interpreted cautiously due to potential unmeasured confounding factors.