The effects of altering the minimum chain length α on the alignment predicted by the CAF algorithm. (A–C) Using the Evolver mammals data set. (D–F) Using the Evolver primates data set. (G–I) Using the Blanchette data set. (A,D,G) Precision-recall plots. The variable α is altered in powers of 2 from 4 to 8192. In general, increasing α increases precision and decreases recall. Four strategies are shown, for the four combinations involving simple melting/progressive melting and simple annealing/progressive annealing. (B,E,H) The F0.25 score (weighting precision fourfold over recall) as a function of α, for the four CAF strategies (colors consistent with [A,D,G]). (C,F,I) The computation time (using a single contemporary 0 × 86 processor) as a function of α.
