The relative contribution of mutation and purifying selection to trans
ition bias has not been quantitatively assessed in mitochondrial prote
in genes. The observed transition/transversion (s/v) ratio is (mu(s),
P-s)/(mu(v) P-v), where mu(s) and mu(v) denote mutation rate of transi
tions and transversions, respectively, and P-s and P-v denote fixation
probabilities of transitions and transversions, respectively. Because
selection against synonymous transitions can be assumed to be roughly
equal to that against synonymous transversions, P-s/P-v approximate t
o 1 at fourfold degenerate sites, so that the s/v ratio at fourfold de
generate sites is approximately mu(s)/mu(v), which is a measure of mut
ational contribution to transition bias. Similarly, the s/mu ratio at
nondegenerate sites is also an estimate of mu(s)/mu(v) if we assume th
at selection against nonsynonymous transitions is roughly equal to tha
t against nonsynonymous transversions. In two mitochondrial genes, cyt
ochrome oxidase subunit I (COI) and cytochrome b (cyt-b) in pocket gop
hers, the s/v ratio is about two at nondegenerate and fourfold degener
ate sites for both the COI and the cyt-b genes. This implies that muta
tion contribution to transition bias is relatively small. In contrast,
the s/v ratio is much greater at twofold degenerate sites, being 48 f
or COI and 30 for cyt-b. Given that the mu(x)/mu(v) ratio is about 2,
the P-s/P-v ratio at twofold degenerate sites must be on the order of
20 or greater. This suggests a great effect of purifying selection on
transition bias in mitochondrial protein genes because transitions are
synonymous and transversions are nonsynonymous at twofold degenerate
sites in mammalian mitochondrial genes. We also found that nonsynonymo
us mutations at twofold degenerate sites are more neutral than nonsyno
nymous mutations at nondegenerate sites, and that the COI gene is subj
ect to stronger purifying selection than is the cyt-b gene. A model is
presented to integrate the effect of purifying selection, condon bias
, DNA repair and GC content on s/v ratio of protein-coding genes.