A methodological bias toward overestimation of molecular evolutionary time scales

Abstract

There is presently a conflict between fossil-and molecular-based evolutionary time scales. Molecular approaches for dating the branches of the tree of life frequently lead to substantially deeper times of divergence than those inferred by paleontologists. The discrepancy between molecular and fossil estimates persists despite the booming growth of sequence data sets, which increasingly feeds the interpretation that molecular estimates are older than stratigraphic dates because of deficiencies in the fossil record. Here we show that molecular time estimates suffer from a methodological handicap, namely that they are asymmetrically bounded random variables, constrained by a nonelastic boundary at the lower end, but not at the higher end of the distribution. This introduces a bias toward an overestimation of time since divergence, which becomes greater as the length of the molecular sequence and the rate of evolution decrease.T he hypothesis of the molecular clock holds that the number of amino acid (or nucleotide) replacements in any given protein (or DNA) sequence changes linearly with time (1, 2). If constant, rates of molecular evolution can be extrapolated for dating past evolutionary events. Rates used for extrapolation have to be first calibrated by reference to absolute dates drawn from the fossil record. A notable feature of the hypothesis of the molecular clock is multiplicity: every one of the thousands of proteins or genes of an organism is an independent clock, each ticking at a different rate, but all measuring the same events (3-5). Molecular clock projections have ostensibly pushed back fossil-based dates in many studies (6). Prominent examples are the time of origin of the metazoan phyla, which has been placed as twice as old as determined by paleontologists (but see ref. 7), dating to more than 1,000 Myr ago (8-10); or the split of the three multicellular kingdoms, timed at about 1,600 Myr ago (9-11), some 400 Myr earlier than predicted from the fossil record.Two not mutually exclusive explanations have been adduced to account for molecular earlier than fossil dates: (i) incompleteness of the fossil record, such that paleontological data can provide only minimal divergence dates (6, 12, 13); and (ii) too few genes and proteins considered, which turns molecular dating methods inaccurate (6, 9). It has been proposed that discrepancies between fossil and molecular dates will fade away as new fossil findings continue to accumulate; but also, and more steeply, as the size of molecular data sets become increasingly larger, because averages across numerous estimates of the same date will converge toward more consistent estimates (6,9,10,14). Yet although data sets have become much larger and methods of analysis considerably more sophisticated, the discrepancy between fossil and molecular dates has not disappeared (reviewed in ref. 6). We now show that common molecular estimates are upwardly biased because of a fundamental f law in the molecular approach to dating.Suppose three orthologous protein sequenc...

Genomic data support the hominoid slowdown and an Early Oligocene estimate for the hominoid–cercopithecoid divergence

“…First, simulations show that early single calibration points can bias clock estimates of older divergence dates upwards (68). In these simulations, the estimated dates begin to converge on the actual dates when the length of the simulated data sets approached 500 aa.…”

Invasion of the continents: cyanobacterial crusts to tree-inhabiting arthropods

“…Continental clades were determined by the author variously by: (i) a preponderance of species that are currently living on land or freshwater; (ii) those that have origins or basal clade synapomorphies that are tied to a significant non-marine existence; (¡ii) having a predominantly non-marine fossil record; (iv) occurrence principally on continental hosts if theterrestrialized clade is parasitic; or (v) combinations of the above. Many early microorganismic dates are not supported by Precambrian (Archean-I-Proterozoic) fossils, although the Cyanobacteria and Archaea are; additionally, dates from molecular clock estimates often are often overestimated based on estimates of calibration points and divergence nodes that lack statistical error as well as methodological biases [50,80,81]. The phylogeny in Figure la does not depict extensive lateral transfer of genes among groups and subgroups; such reticulate evolution was significant during the Precambrian [5,75,76,82], and explains the absence consensus among comparisons of trees from different gene systems [82].…”

Estimating metazoan divergence times with a molecular clock

“…Because molecular clocks have several inherent problems, including how the clock is calibrated, how molecular substitution rates are estimated, and how heterogeneity in these rates is detected and corrected (3,4), as well as an inherent statistical bias for overestimating dates (4,17), a much more recent date for LCB may not yet be refuted. Of crucial importance for clock accuracy is the calibration of the clock itself, which requires not only accurate paleontological estimates (18) but also rate homogeneity between the calibrated and uncalibrated taxa.…”