Table 6. Positive selection on single codon level based on PAML (see text for details).

Genea	Modelsb	ωSc	p(ωS)d	−2ΔLLb	Pb	aae
GNBP	M1 vs. M2	1	9.1	0	Ns	None
GNBP	M7 vs. M8	1	3.8	5.9	Ns	None
SP14D1	M1 vs. M2	1	1.1	0	Ns	None
SP14D1	M7 vs. M8	1.97	2.4	6.6	0.037	206ns; 169ns
Gambicin	M1 vs. M2	12.1	1.3	6.4	0.041	72**
Gambicin	M7 vs. M8	11.1	1.3	11.8	0.001	72**
Defensin	M1 vs. M2	1.4	0	1.1	Ns	None
Defensin	M7 vs. M8	2.2	2.4	0.9	Ns	None

^a GNBP alignment was 171 aa long and included eight species; SP14D1alignment was 246 aa long and included six species; Gambicin alignment was 81 aa long and included nine species; Defensin alignment was 101 aa long and included seven species (see Materials and Methods for the species listing for each gene).

^bLikelihood ratio tests (with 2 df) were used to determine the significance of finding ω>1 over all codons by comparing selection models (M2 and M8) that allowed for ω>1 with neutral (M1 and M7) models that allowed only ω≤1.

^cEstimate of the highest ω value for any codon.

^dThe proportion of codons with the highest w estimate.

^ePositions of the amino acids with ω>1 and their significant value estimated by BEB test in PAML.