Maximum likelihood estimates of dN and dS using codon-based models.

1)Assuming that M0 is true null hypothesis (Η0) with p0 parameters and that M3 is the alternative hypothesis (Η1) with p1 parameters. The log likelihood difference 2Δℓ = 2(ℓ1−ℓ0) follows asymptotically a χ2 distribution with (p1−p0) degrees of freedom.*)Best-fitted model according to the likelihood ratio test (LRT) between of M0 vs. M3 with 4 degrees of freedom.$)Best-fitted model according to the likelihood ratio test (LRT) between of M2 vs. M3 with 1 degree of freedom.#)Positively selected sites were detected according to the Naive Empirical Bayes method with p≥0.99. Sites 33 and 44 (underlined) correspond respectively to sites 11 and 25 of V3 loop; which are related with R5-to-X4 change.∶)This result was obtained using unique sequences and X4-strains because exclusion of them did not affect on the model preference or the parameter values.