Skip to main content
Fig. 4 | Journal of Cheminformatics

Fig. 4

From: Detecting drug communities and predicting comprehensive drug–drug interactions via balance regularized semi-nonnegative matrix factorization

Fig. 4

Community images of the DDI adjacent matrix. a Original DDI matrix, b rearranged DDI matrix upon three communities generated by Semi-NMF; c rearranged DDI matrix upon three communities generated by BRSNMF. Each pixel in the DDI matrix image represents a drug pair. Magenta pixels represent enhancive DDIs, cyan pixels represent degressive DDIs and black pixels are non-DDIs. The boundaries of communities are highlighted by white lines. The numbers of drugs in communities in all the images are exhibited as well. Semi-NMF generates the communities with larger variance of community sizes and cannot detect the weakly balanced community. In contrast, BRSNMF generates the communities which are less dispersed in terms of community size. More importantly, BRSNMF is able to detect a weakly balanced community

Back to article page