Predicting disease associations via biological network analysis

Table 1 Comparison of studies on inferring disease-disease associations

	Data	Size	Evaluation
van Driel et al. (2006) [4]	OMIM	5132 phenotypes in OMIM	Comparing results with genotypic similarities
Lage et al. (2007) [5]	OMIM	7000 OMIM record pairs	Evaluating results against the overlap of the OMIM record pairs
Goh et al. (2007) [6]	OMIM	1284 OMIM diseases	Analysing network topologicalproperties
Huang et al. (2009) [12]	GWAS	7 diseases	Comparing results with phenotypic similarities
Li and Agarwal (2009) [7]	Pubmed abstracts,biological pathways	1028 diseases in MeSH	Comparing results with MeSHclassification
Kim et al. (2009) [13]	GWAS	53 clinical traits related tosevere asthma	Mining the literature manually
Hu and Agarwal (2009) [8]	Expression data	645 diseases in MeSH	Comparing results with MeSHclassification
Suthram et al. (2010) [9]	Expression data, PPI	54 diseases	Evaluating results against genetic similarities
Lewis et al. (2011) [14]	GWAS	61 diseases	Comparing results with Huang et al.(2009) results
Mathur and Dinakarpandian et al.(2007) [10]	DO annotation, GOannotation	36 diseases (for evaluation)	Evaluating results using 68 curated disease associations
Our study	Disease-gene associations, GOannotation, PPI	543 ICD-9 diseases	Evaluating results against ICD-9classification, comorbidity, andgenetic similarities derived fromGWAS data

The comparison is based on the data used to derive associations (denoted by ‘Data’), number of diseases evaluated (denoted by ‘Size’) and benchmarks used for evaluation (denoted by ‘Evaluation’). The number of diseases evaluated in our study is computed as the union of diseases annotated in the four disease-gene association datasets we analysed, given in Figure 1.