Table 3.

Performance on the CAGI 13 dataset.

Method	Ranking Details	AUC
Uniform weight 1	Count SNVs in CD genes.	0.743
Uniform weight 2	Count SNVs in CD genes and their STRINGdb interactors.	0.736
Mixed pedigree 1	Max one control per family. Count SNVs in CD genes.	0.844
Mixed pedigree 2	Max one control per family. Count pathogenic SNVs (according to SIFT) in CD genes.	0.688
Mixed pedigree 3	Max one control per family. Look for families with large difference in SNVs count in CD genes.	0.798
Bi-clustering	Bi-clustering on the dataset. Rank according to the number of SNVs in CD genes.	0.866

The submissions have a high correlation, ranging between 0.67 and 0.89. This is mainly due to (1) a similar set of variants selected for all methods and (2) the use of transductive clustering from CAGI 11. The use of clustering is critical to maximize performance.