Fig. 8From: BioBLP: a modular framework for learning on multimodal biomedical knowledge graphsDifference in macro MRR between RotatE and BioBLP-D when a disease is predicted, as we consider increasing values of node degree. We observe that when considering nodes with a degree of 74 or less, BioBLP-D results in improved link prediction performance (shown as negative values of \(\Delta\)MRR)Back to article page