Figure 3
Clinical trial success and failure prediction. (A) Distribution of Rosalind scores for true failure versus true success when predicting on the relation Clinical Trial Failure, (B) when predicting on the relation Clinical Trial Success, and (C) when predicting on the benchmark relation, Therapeutic Relationship. For boxplots, white horizontal lines indicate the median of the distributions. The box extends between the first and third quartile (and show the interquartile range). Whiskers extend to the ends of the distribution. Plots do not show outliers. Shown below A, B and C are the histograms of the scores for clinical success and failure when predicting on the relation shown in the boxplot above. Significance was tested using an MWW test. P-values are Bonferroni-corrected for multiple comparisons, resulting in a significance threshold of \(\alpha\)=0.017. ***\(p<\)0.001. (D) Clinical success and failure scores across the MeSH disease tree when predicting on clinical failure. (E) Clinical success and failure scores across the MeSH disease tree when predicting on clinical success.
