Table 2.

Comparison of the average BEDROC scores calculated with various α, for GeneHouds and Endeavour

Methods	Averaged BEDROC score
α	$\alpha =228.5$	$\alpha =160.9$	$\alpha =32.2$	$\alpha =16.1$	$\alpha =5.3$
Early enrichment focus	TOP 100	TOP 1%	TOP 5%	TOP 10%	TOP 30%
$GeneHound\_25LatDims$	0.358 ± 0.057	0.396 ± 0.06	0.595 ± 0.056	0.686 ± 0.045	0.818 ± 0.023
$GeneHound\_30LatDims$	0.36 ± 0.016	0.40 ± 0.017	0.608 ± 0.028	0.70 ± 0.029	0.829 ± 0.019
$GeneHound\_40LatDims$	0.383 ± 0.023	0.421 ± 0.024	0.614 ± 0.027	0.699 ± 0.026	0.824 ± 0.017
$GeneHound\_GeoAgg$	0.418 ± 0.031	0.458 ± 0.031	0.651 ± 0.031	0.733 ± 0.029	0.85 ± 0.018
Endeavour	0.387 ± 0.031	0.422 ± 0.030	0.609 ± 0.029	0.694 ± 0.029	0.817 ± 0.021

Notes: The confidence intervals are reported over folds. α tunes the early enrichment. For example, in our study when $\alpha =228.5$, 80% of BEDROC score is given to the top 100 ranked genes. The best performance for each α is shown in boldface. All models are benchmarked on OMIM2.

The maximum BEDROC scores at different early enrichment focuses are highlighted in bold.