Skip to main content
Fig. 1 | BMC Bioinformatics

Fig. 1

From: AttentionDDI: Siamese attention-based deep learning method for drug–drug interaction predictions

Fig. 1

Modality importance using attention scores and masking methods for DS1. First and second rows report modality importance using the masking approach (see Algorithm 1). The values represent the average models’ relative change in AUC and AUPR performance when masking applied to each modality one at a time compared to a base model that has access to all modalities. Third row represents the average between AUC and AUPR average relative change values (i.e. average of values in first and second rows). Fourth row reports average modality importance using Attention score computation (see Eq. 21). The higher the value, the more important the data modality is

Back to article page