Fig. 2

Comparative or MFE? As in Fig. 1, each panel depicts the ROC performance of a classifier based on thresholding the T-S ambiguity index, with small values of d_T-S(p,s) taken as evidence that s was derived by comparative as opposed to MFE secondary structure analysis. Left Panel: performance on molecules chosen from the unbound group. Right Panel: performance on molecules chosen from the bound group. Conditional p-values were also calculated, using the hypergeometric distribution and based only on the signs of the indexes. In each case the null hypothesis is that comparative secondary structures are as likely to lead to positive ambiguity indexes as are MFE structures, whereas the alternative is that positive ambiguity indexes are more typical when derived from MFE structures. Left Panel: p=5.4×10⁻¹⁴. Right Panel: p=0.07