Skip to main content

Table 6 Results on KIBA and Davis of our proposed multi-granularity multi-scaled SANs model, transitional methods and existing deep sequence representation methods

From: Multi-scaled self-attention for drug–target interaction prediction based on multi-granularity representation

 

Method

Drug

Protein

Interaction

CI

MSE

\(r^2_m\)

KIBA

KronRLS [16]

Pubchem Sim

S-W

0.782 (0.001)

0.411

0.342 (0.001)

SimBoost [17]

Pubchem Sim

S-W

0.836 (0.001)

0.222

0.629 (0.007)

DeepDTA [6]

CNNs

CNNs

Concatenation

0.863 (0.002)

0.194

0.673 (0.009)

MT-DTI [23]

SANs

SANs

Concatenation

0.882 (0.002)

0.152

0.738 (0.006)

GANsDTA [20]

GANs

GANs

Concatenation

0.866 (−)

0.224

0.675 (−)

CrossAttentionDTI [24]

Cross SANs

Cross SANs

Concatenation

0.874 (0.001)

0.175

Ours

MSSAN

MSSAN

Concatenation

0.890 (0.002)

0.155

0.742(0.010)

Davis

KronRLS [16]

Pubchem Sim

S-W

0.871 (0.001)

0.379

0.407 (0.005)

SimBoost [17]

Pubchem Sim

S-W

0.872 (0.001)

0.282

0.644 (0.006)

DeepDTA [6]

CNNs

CNNs

Concatenation

0.878 (0.004)

0.261

0.630 (0.017)

MT-DTI [23]

SANs

SANs

Concatenation

0.887 (0.003)

0.245

0.665 (0.014)

GANsDTA [20]

GANs

GANs

Concatenation

0.881 (−)

0.276

0.653 (−)

CrossAttentionDTI [24]

Cross SANs

Cross SANs

Concatenation

0.876 (0.006)

0.244

Ours

MSSAN

MSSAN

Concatenation

0.890 (0.005)

0.233

0.681 (0.014)

  1. Bold values indicate the best results on the datasets