Fig. 4
From: Estimating statistical significance of local protein profile-profile alignments
Statistical accuracy of the COMER method using different models for statistical significance estimation. Shown are the results of comparing the profiles constructed for 5000 randomized Pfam families to the profiles constructed for 4931 randomized SCOPe domains. The figure plots the fraction of queries with a p-value reported for their best match less than the p-value indicated on the x-axis. The solid straight line indicates the highest accuracy. These models are represented in the figure: the new model (new) proposed, the alternative model with E-value correction (new (alt.)), a statistical model based on previous research (S&G’08), the statistical model used in the previous COMER version (previous), a model based on the statistics obtained for profiles generated with s=1 and r=0.01, and the model excluding the number of positive substitution scores (w/o ωn). The statistical accuracy of HHsearch and COMPASS (v3.1) is also shown