2.5.3 Tests for Selection
For selection, we conduct a priori classification of peptide binding region (PBR) and non- peptide region upon inferred passerine PBR sequences [40, 41] homology sites with chicken MHC [42, 43] and human HLA [44]. The identification of sites subjected to selection in MHC class I Exon 3 was performed using various methods. The first standard selection test (Tajima’s D, Fu & Li’s F* and Fu & Li’s D*) were calculated using DnaSP 5.0 [32]. Second method was the calculation of parameter (ω) for functional alleles. It was carried out an overall estimation ofd N/d S of MHC class I Exon 3 and the other was codons comprising only PBR and non-PBR, which was calculated with MEGA 7.0 according to the Nei-Gojobori method [45] with the Jukes and Cantor correction. Standard error estimates were derived from 1000 bootstrap replicates. Z test of historical positive selection [46] was calculated in MEGA 7.0. Third, the Maximum likelihood implemented in codeml in PAML 4.9 was used for identification of sites involved in the positive selection, which are indicated where the ratio ω (d N/dS ) larger than 1 [47]. Two different models corresponding ω were tested: M7 (beta), M8 (β and ω). To find whether the alternative model (M8) provided better fitter than the M7, we performed Likelihood ratio tests to compare twice the difference of the log-likelihood ratios (2ΔlnL) using a distributionχ 2. PSSs in the M8 model was identified by PP more than 95% using the Bayes empirical Bayes procedure. Positively selected sites were verified at each codon site separately using many complementary approaches implemented in datamonkey (http://www.datamonkey.org/) [48] in addition to afore mention methods. Specifically, we used MEME [49], FEL, SALC [50] and FUBER [51]