A molecular clock is a method for estimating when evolutionary lineages diverged by relating differences in DNA, RNA, or protein sequences to the accumulation of evolutionary changes over time. It is a central tool in phylogenetics, allowing evolutionary relationships to be placed on a timescale. The term also refers to the hypothesis that particular molecular sequences accumulate changes at an approximately constant rate. Modern methods need not assume a universal constant rate: they can accommodate differences among lineages, genes, and sites within a sequence.
Basic principle
As lineages evolve, inherited molecular changes accumulate along their branches of an evolutionary tree. Comparisons of homologous sequences—sequences derived from a shared ancestral sequence—provide information about the amount of evolutionary change separating them. If the rate of change can be estimated, that amount can be translated into elapsed time.
In the simplest case, two sequences diverge from a common ancestral sequence and subsequently evolve at the same constant rate:
where:
- is the expected number of substitutions per sequence site separating the sequences;
- is the substitution rate per site per unit of time along each lineage;
- is the time since their common ancestral sequence.
The factor of two reflects change on both descendant branches. If the two lineages have different constant rates, the corresponding expression is .
Here, a substitution is an evolutionary replacement represented along a lineage, not simply any newly arising mutation. Many mutations disappear without contributing to long-term differences between lineages.
The observed proportion of differing sites is not generally equal to . A site may change more than once, return to an earlier state, or independently acquire the same state in different lineages. A nucleotide substitution model estimates the otherwise hidden changes. Without such correction, sequence differences tend to underestimate evolutionary change, especially over long timescales.
Molecular data alone usually identify the product of rate and time more readily than either quantity separately. A branch with many changes could represent a long interval at a slow rate or a short interval at a fast rate. Estimating absolute dates therefore requires external temporal information or an independently justified rate.
Historical development
The molecular-clock concept emerged from comparisons of protein sequences during the early 1960s. Emile Zuckerkandl and Linus Pauling observed that differences among related proteins, including hemoglobin, could provide information about the timing of evolutionary divergence. Their work in 1962 and 1965 helped establish molecular sequence change as a potential measure of evolutionary time.
The proposal was influential because molecular evidence offered a source of chronological information independent of anatomical comparisons. It also raised a theoretical question: why should molecular evolution show any regularity when organisms differ greatly in their environments, life histories, and patterns of adaptation?
The neutral theory of molecular evolution, developed by Motoo Kimura from 1968, supplied an important explanation. Subsequent work showed that exact rate constancy is uncommon across broad groups, encouraging the development of local and relaxed molecular-clock models.
Biological basis of clock-like evolution
Under a simple neutral model, the long-term substitution rate at neutral sites equals the neutral mutation rate.
For an idealized diploid population with effective size , approximately new neutral mutations arise per generation at a site, where is the mutation rate per copy per generation. Each new neutral mutation has fixation probability . Their product gives the expected substitution rate:
This result connects molecular evolution to genetic drift and explains why a roughly steady mutation process can generate an approximately steady substitution process.
It does not imply that all molecular changes are neutral or that all clocks run at the same speed. Natural selection removes many harmful changes and can promote advantageous ones. Functional constraints differ among genes and among sites within a gene. For example, an amino acid essential to a protein’s activity may change much less frequently than a weakly constrained position.
Rates also depend on mutation mechanisms. Differences in DNA replication, DNA repair, generation time, and other biological characteristics can produce different rates among lineages. A constant rate per generation does not necessarily imply a constant rate per year.
Main clock models
Strict clocks
A strict molecular clock assigns the same substitution rate to every branch of a tree for a specified sequence or data partition. Rates may still differ among sites or among separately modeled genes.
For sequences sampled at the same time, a strict clock implies equal expected evolutionary distances from the root to every sampled tip. Such a tree is called ultrametric. Observed sequence differences nevertheless fluctuate because substitutions are stochastic and sequences are finite.
Strict clocks can be effective when rate variation is small. If substantial variation is ignored, inferred divergence times may be biased.
Local clocks
Local-clock models assign one rate to a specified group of branches and different rates to other groups. They are useful when a rate shift is concentrated in a particular lineage or clade.
Some methods require the rate groups to be specified in advance; others estimate their placement from the data.
Relaxed clocks
Relaxed-clock models allow branch rates to vary. Important types include:
- Autocorrelated models, in which descendant rates tend to resemble ancestral rates.
- Uncorrelated models, in which branch rates are drawn independently from a shared probability distribution.
- Random local-clock models, in which rate changes occur on a subset of branches inferred during analysis.
These models provide flexibility but introduce additional parameters and assumptions. A relaxed clock does not automatically eliminate dating error: its performance depends on the data, the calibration information, and whether its assumptions adequately describe the evolutionary process.
Calibrating evolutionary time
A molecular clock requires a connection between genetic change and calendar time. Several kinds of evidence can supply that connection.
Fossils. Evidence from paleontology can constrain the ages of lineages. A securely assigned fossil generally establishes that the relevant lineage existed by the fossil’s age, providing a minimum constraint rather than an exact divergence date. Maximum constraints require additional justification because the absence of older fossils does not by itself demonstrate that a lineage was absent.
Dated samples. Sequences collected at different known times can reveal an evolutionary rate directly when enough change has accumulated across the sampling interval. This approach is especially useful for rapidly evolving viruses. Dated ancient DNA can extend the temporal range for other organisms.
Geological or biogeographic events. An independently dated event, such as island formation or the emergence of a geographic barrier, can sometimes inform divergence times. The inference requires caution: a lineage’s split or colonization need not coincide with the event.
External rate estimates. Rates estimated from pedigrees, experiments, or other evolutionary studies may provide calibration information. Their relevance depends on the organism, genomic region, timescale, and definition of the measured rate.
Calibration uncertainty is commonly represented by bounds or probability distributions rather than fixed dates. Multiple independent, well-supported calibrations can strengthen an analysis, but incorrectly assigned or incompatible constraints can distort the entire timescale.
Molecular dating differs from radiometric dating. It measures biological sequence change rather than radioactive decay, although radiometric dates may establish the ages of fossils or associated geological deposits used as calibrations.
Statistical inference and uncertainty
A typical analysis combines sequence data, a tree model, a substitution model, a clock model, and temporal constraints. Maximum-likelihood and Bayesian inference methods are widely used.
Bayesian analyses combine the likelihood of the observed sequences with prior distributions on rates, divergence times, and other parameters. They produce posterior distributions rather than a single uniquely determined chronology. When the molecular data contain little temporal information, the priors and calibrations can strongly influence the results.
Important sources of uncertainty include:
- limited sequence length and incomplete taxon sampling;
- errors in sequence alignment or reconstructed relationships;
- inadequately modeled rate variation;
- uncertainty in fossil placement and age;
- assumptions about lineage diversification and sampling;
- conflicting evolutionary histories among genomic regions.
Reported intervals are conditional on the model and calibration choices. A narrow interval does not guarantee accuracy if important assumptions are wrong. Sensitivity analyses, alternative calibrations, and tests of temporal signal help assess how strongly the conclusions depend on those choices.
Applications
Molecular clocks provide timescales for major events in evolution, including the diversification of organisms and the origins of particular groups. They help compare evolutionary histories with changes in geography, climate, and the fossil record.
In human evolution and population genetics, molecular dating contributes to estimates of ancestral relationships and population history. A genetic date must nevertheless be interpreted according to what was modeled: the common ancestry of sampled sequences is not automatically the date when populations separated.
In epidemiology, dated pathogen sequences can help reconstruct the timing of lineage emergence and spread. The estimated common ancestor of sampled pathogen sequences is not necessarily the first infection or the beginning of an outbreak.
Different questions require different molecular markers. Slowly evolving sequences can retain information over deep timescales, while faster-evolving regions may distinguish recent events.
Limitations and interpretive issues
Gene divergence is not species divergence
Copies of a gene sampled from different species can share a common ancestor older than the separation of those species. Ancestral genetic variation may persist through successive splitting events, causing gene trees to differ from species trees.
Coalescent theory describes this backward-in-time ancestry of gene copies. Multispecies coalescent models explicitly distinguish gene histories from species divergence.
Saturation and model misspecification
Over long intervals, repeated substitutions can erase evidence of earlier changes. This substitution saturation limits the chronological information retained in a sequence. Incorrect assumptions about sequence composition, selective constraints, or changes in the substitution process can further bias estimates.
More sequence data reduce sampling error, but do not necessarily correct systematic errors in the model.
Recombination and horizontal transfer
Recombination can give different parts of a sequence different genealogies. Horizontal gene transfer can similarly make a gene’s history differ from that of its host organisms. Treating such data as if they evolved along one branching tree can produce misleading dates.
Timescale-dependent rate estimates
Rates inferred over short intervals can exceed those estimated over long evolutionary periods. Contributing mechanisms include the temporary presence of mildly harmful variants that are later removed by selection, unrecognized multiple substitutions, and differences in calibration or sampling.
Consequently, a short-term mutation or evolutionary-rate estimate should not be assumed to apply unchanged to deep divergence times. Molecular-clock estimates are conditional reconstructions whose interpretation depends on the evolutionary process, the timescale, and the evidence used to calibrate them.