TABLE 3

ANOVA significance tests of effects on domain-specific KA/S values

SourceSSMSDFFP
All sequences
Tissue0.2480.24813.610.072
Gene[tissue]{random}1.310.082162.990.0005
Domain0.8990.11284.100.0003
Domain*Tissue0.5290.06682.410.021
Restricted sample (KS ≤ 0.2)
Tissue0.3970.39716.170.027
Gene[tissue]{random}0.8350.076112.940.0034
Domain0.5190.06582.520.019
Domain*Tissue0.6050.07682.930.0076

KA was estimated separately for each domain of each pair of orthologous OR sequences and then normalized by the overall estimate of KS for that gene; these normalized nonsynonymous substitution rates were used as the dependent (response) variable. The independent variables were tissue (nose or testis), domain, and gene identity (which was nested within tissues and treated as a random effect). The very short and relatively poorly represented IC1 domain was omitted. The analysis was performed both with and without the five genes showing KS values >0.2. Not surprisingly, genes and domains show strong and significant main effects in both analyses. The most interesting result is the significant domain*tissue interaction, which also appears in both analyses. This indicates that the NOR and TOR samples show distinct patterns of differences (in levels of amino acid conservation) among domains of the OR protein (see Table 2). Owing to the significance of this interaction effect, the main effect of tissue has no simple interpretation and therefore should not be taken at face value.