Number of sites falsely identified as having evolved under positive selection

MethodCutoffWrong sites%Range (%)
SLR0.95 1529 52–43 (1–14)
0.99  334 10–14 (0–5)
SNY0.9510646350–300 (0–100)
0.99 9540320–300 (0–100)
SNY + LRT950.95 2808 90–300 (0–100)
 (17)0.99 2586 90–300 (0–100)
SNY + LRT990.95 1171 40–300 (0–100)
 (5)0.99 1106 40–300 (0–100)
  • One hundred data sets, each 300 codons long, were simulated under a model of neutral evolution and analyzed. “SNY + LRTx (n)” refers to detecting sites using empirical Bayes on data sets that pass a statistical test for the presence of positive selection using significance level x%, the number n in parentheses being the number of such data sets (out of 100) that passed the test. “Cutoff” refers to the significance level used in LRTs for the SLR test and to the empirical Bayes posterior probability cutoff level used in the SNY test—note that these numbers are not directly comparable (see text for further details). “Range” shows the minimum and maximum numbers of falsely identified sites observed within the 100 individual data sets.