Skip to main content

Table 3 Evaluation of our measures against comorbidity

From: Predicting disease associations via biological network analysis

Data

Annotation-based

Function-based

Topology-based

OMIM

0.8009 ± 0.0277 (0.5740)

0.8694 ± 0.0073 (0.5120)

0.8495 ± 0.0011 (0.5044)

CTD

0.7849 ± 0.0164 (0.5404)

0.7316 ± 0.0046 (0.5047)

0.7949 ± 0.0042 (0.5203)

FunDO

0.7426 ± 0.0088 (0.4672)

0.7142 ± 0.0017 (0.4940)

0.7497 ± 0.0016 (0.5031)

HuGENet

0.7563 ± 0.0001 (0.5084)

0.8185 ± 0.0001 (0.4987)

0.7153 ± 0.0015 (0.4922)

Intersection

0.9925 ± 0.0001 (0.6013)

0.9802 ± 0.0001 (0.5081)

0.9958 ± 0.0041 (0.4664)

Union

0.8225 ± 0.0045 (0.4704)

0.7491 ± 0.0001 (0.4999)

0.7939 ± 0.0022 (0.5008)

Average

0.8194 ± 0.0837 (0.5270)

0.8106 ± 0.0930 (0.5029)

0.8163 ± 0.0907 (0.4979)

  1. Numbers in the table are AUC values obtained by evaluating the three disease similarity measures against comorbidity associations. The ϕ-correlation threshold was set to 0.06 (the same threshold was used in [47]), and all diseases annotated with at least 3 genes were evaluated. Average AUC values obtained by using randomised scores are shown by numbers in brackets (standard deviations are not shown in the table due to space limitation). Each evaluation test was run 30 times to compute the statistics reported in the table.