Table 2.

Estimates of the mutation, selection and demographic parameters for introns and synonymous sites. (N_a and N_b are the effective population sizes after and before the population expansion; g = N_a/N_b; τ is the time since the expansion (in units of 2N_a generations); κ is the mutational bias; γ is the equivalent of the selection coefficient in favour of heterozygotes at a site, multiplied by 4N_a. L₀ is a model with selection on GC versus AT basepairs at intronic sites and P versus U codons at synonymous sites, but no population expansion. The full L₁ model is the same as model L₀ with population expansion; the other L₁ models all have one parameter different from L₁; the last L₁ model has two estimates for γ_int—the smaller value is for introns with low GC content and the larger value is for introns with high GC content. The p-values correspond to the likelihood-ratio test of each alternative model against the full L₁ model, and the ln L rank gives the rank of the log-likelihood among the eight models considered, where 1 indicates the most likely model.)

model	g(Na/Nb)	τ(t/2Na)	γcod	κcod P > U	γint	κint GC > AT	ln L	p-value	ln L rank
L0	—	—	1.89	5.27	0.42	2.43	−12770.70	<0.0001	8
L1 (γcod = 0)	4.49	0.10	0	0.85	0.36	2.31	−12749.21	<0.0001	7
L1 (κcod = 1)	4.59	0.09	0.19	1	0.36	2.31	−12 745.88	<0.0001	6
L1 (κint = 1)	18.40	0.49	3.02	20.42	−0.46	1	−12743.26	<0.0001	5
L1 (γcod = γint)	4.40	0.10	0.55	1.45	0.55	2.77	−12741.58	<0.0001	4
L1 (γint = 0)	4.25	0.11	1.55	3.87	0	1.65	−12738.40	0.004	3
full L1	4.21	0.11	1.56	3.87	0.36	2.31	−12734.27	—	2
L1 (γint low GC, γint high GC)	4.20	0.11	1.55	3.87	0.27, 0.45	2.32	−12724.84	<0.0001	1