Skip to main content

Table 4 Results of ablation test on DeepMPF for AUC

From: DeepMPF: deep learning framework for predicting drug–target interactions based on multi-modal representation with meta-path semantic analysis

Dataset

DeepMPF (ours)

Without sequence modality (-)

Without heterogeneous structure modality (-)

Without similarity modality (-)

Only modality of sequence (-)

Enzyme

0.9645

0.9618 (− 0.0027)

0.9405 (− 0.0240)

0.9503 (− 0.0142)

0.8462 (− 0.1183)

GPCR

0.8781

0.8628 (− 0.0153)

0.8403 (− 0.0378)

0.8573 (− 0.0208)

0.7599 (− 0.1182)

Ion channel

0.9762

0.9747 (− 0.0015)

0.9463 (− 0.0299)

0.9657 (− 0.0105)

0.7341 (− 0.2421)

Nuclear receptor

0.8271

0.7870 (− 0.0401)

0.8131 (− 0.0140)

0.7938 (− 0.0333)

0.8050 (− 0.0221)

  1. The bold values represent the maximum drop value of AUC on each dataset among the comparisons except the term of sequence only