identify and exclude sequences that violate the rate-constancy assumption have only
limited statistical power (Dobzhansky et al. 1977; Scherer 1989; Robinson et al. 1998;
Bromham et al. 2000). Dramatic evolutionary rate differences among lineages can pass
undetected by conventional molecular clock tests for most common alignment lengths.
The possibility of deriving extremely erroneous conclusions on the basis of statistically
insignificant rate variation is illustrated by GPDH and SOD.
Be that as it may, there are important evolutionary questions, such as the configuration
of a ‘universal tree’ of life (Woese et al. 1990), for which gene and protein sequences may
provide the best, if not the only dependable information. Our own recommendation is
rather pat: (1) Use molecular sequence data not as definitive, as sometimes is done (e.g.
Wray et al. 1996; see Ayala et al. 1999), but as one more source of information when
other evidence is available. (2) Pay attention to careful choice of the sequences. Close
approximation to the molecular clock premise should be a necessary condition. Given the
limited power of available tests, however, acceptance of this second premise seems safe
only for long and fast evolving (yet alignable) sequences. (3) Combine data for as many
genes as feasible, so that average values may converge towards a good correlation between
amount of change and time elapsed—the so-called ‘law of large numbers’.
Table 1.4 shows two sets of evolutionary rates and clock estimates derived from these
rates. The evolutionary rates are based on the average rates given in Table 1.3 (last
column) and are normalized to the average values for comparisons between (1) Drosophila
subgenera and (2) mammal orders. Time estimates derived from (1) underestimate the
time of divergence of mammals, animal phyla and kingdoms. This underestimation occurs
because of the relatively fast rate of evolution of the Drosophila lineages for most genes
(see Figure 1.3 and Table 1.4), with the notable exception of GPDH, which is not
sufficient to overcome the effect of the other genes. Time estimates derived from (2) would
seem more nearly accurate for the animal phyla and kingdoms, but overestimate the time
of divergence of the Drosophila subgenera. In any case, the time estimates derived from the
average rates are, as expected, generally more accurate than those that would be obtained
Table 1.4 Normalized rates of evolution across the loci, normalized to the rates of Drosophila
subgenera and mammal orders, and corresponding estimates of divergence times
The normalized rates are derived from the averages in Table 1.3: (1) normalized to the Drosophila-
subgenera average; (2) normalized to the mammal-orders average. The clock estimates assume that
the divergence times are 60 Ma for the Drosophila subgenera and 70 Ma for the mammal orders.
22 FRANCISCO RODRÍGUEZ-TRELLES ET AL.