Fig. 4From: Predicting clinically promising therapeutic hypotheses using tensor factorizationValidation of BTF model prediction. a t-SNE visualization of indications based on the latent factors learned in BTF model. Each dot represents one indication and the size of the dot is proportional to the number of targets that have been clinically failed. The inserted pie charts show diseases composition of representative clusters of indications in the 2D visualization. 2D embedding was obtained by using perplexity = 30 in t-SNE and the visualization is consistent using different perplexity values in the range from 10 to 50. b BTF prediction scores of target-indication pairs in Phase I-III clinical trials. The numbers are P-values (Wilcoxon rank sum tests) from comparing prediction scores of target-indication pairs between any two phasesBack to article page