Skip to main content

Advertisement

Table 2 Site models results: estimated parameters and log-likelihood ratio tests

From: Contrasted patterns of selective pressure in three recent paralogous gene pairs in the Medicagogenus (L.)

Gene Model np logL Parameters LRT p-value
Lax2 M8A 35 −1600.22 p = 0.01 q = 2.86    
     ωad = 1 pad = 0.03    
  M8 36 −1600.19 p = 0.01 q = 3.00 vs. M8A 0.04 0.84
     ωad = 1.10 pad = 0.03    
Lax4 M8A 23 −1391.38 p = 6.30 q = 99.00    
     ωad = 1 pad = 0.00    
  M8 24 −1391.38 p = 6.30 q = 99.00 vs. M8A 0.00 1
     ωad = 1.00 pad = 0.00    
Pg3 M8A 23 −1616.24 p = 4.89 q = 99.0    
     ωad = 1.00 pad = 0.24    
  M8 24 −1615.18 p = 0.13 q = 0.40 vs. M8A 2.11 0.15
     ωad = 4.88 pad = 0.01    
Pg11 M8A 39 −2221.99 p = 2.16 q = 99.00    
     ωad = 1 pad = 0.36    
  M8 40 −2210.38 p = 0.01 q = 20.01; vs. M8A 23.22** 1.44 10-6
     ωad = 6.30 pad = 0.04    
Pg11a M8A 19 −1388.97 p = 2.32 q = 89.36    
     ωad = 1.00 pad = 0.35    
  M8 20 −1385.65 p = 0.47 q = 0.96 vs. M8A 6.64** 9.99 10-3
     ωad = 11.61 pad = 0.02    
Pg11c M8A 21 −1519.41 p = 0.01 q = 2.54    
     ωad = 1.00 pad = 0.32    
  M8 22 −1512.06 p = 0.01 q = 0.05 vs. M8A 14.69** 1.26 10-4
     ωad = 4.45 pad = 0.10    
  1. Note. Models are M8A, ω following a β distribution discretized into 10 categories of similar frequency (0 < ω1–10 < 1) plus an additional category of ωad = 1, accounting for neutral sites; M8 differs from M8A only by the additional category of ω which is constraint to be superior to 1 (ωad > 1), to account for sites under positive selection.
  2. Parameters are frequencies and values of ω for M8A and M8, p and q are the parameters in β distribution; for M8A and M8 pad and ωad are the frequencies and values of additional class of ω; NB ωad is fixed equal to one in M8A. np: number of free parameters; logL: log-likelihood; LRT: likelihood ratio test statistic between indicated models; one (respectively two) asterisk indicates that the probability of observing such an LRT or higher under the compared model is <0.05 (respectively <0.01), assuming that the LRT follows a χ2 distribution with the difference of free parameters between the compared models as the number of degrees of freedom.