Skip to main content

Table 2 Top algorithms by AUC, MRR, and A@10

From: Evaluating semantic similarity methods for comparison of text-derived phenotype profiles

By

Method

AUC

MRR-0

A@10

AUC

AVG (GW) + Resnik (PW) + Resnik (IC)

0.81 (0.79–0.83)

0.19

0.34 (0.31–0.37)

AVG (GW) + Jaccard (PW) + Zhou (IC)

0.8 (0.78–0.82)

0.19

0.32 (0.29–0.35)

AVG (GW) + NODE_SIM (PW) + Zhou (IC)

0.8 (0.78–0.82)

0.18

0.31 (0.28–0.34)

MRR

GIC (DGW) + Zhou (IC)

0.68 (0.65–0.71)

0.24

0.41 (0.38–0.44)

GIC (DGW) + Seco (IC)

0.68 (0.65–0.71)

0.24

0.41 (0.38–0.44)

Bader (DGW)

0.77 (0.74–0.8)

0.24

0.4 (0.37–0.43)

A@10

GIC (DGW) + Resnik (IC)

0.68 (0.65–0.71)

0.24

0.42 (0.39–0.45)

GIC (DGW) + Sanchez (IC)

0.67 (0.64–0.7)

0.24

0.42 (0.39–0.45)

GIC (DGW) + Min (IC)

0.67 (0.64–0.7)

0.23

0.42 (0.39–0.45)

  1. Since MRR-NA and MRR-0 are statically dependent, the top algorithms for both are equivalent, and so only ‘MRR-0’ is listed here