# Molecular Clock Hypothesis: Mechanisms, Methods, and Pitfalls ## Introduction to the Molecular Clock Hypothesis The molecular clock hypothesis posits that the rate of molecular evolution—specifically, the accumulation of nucleotide or amino acid substitutions—is approximately constant over time for a given gene or protein across different lineages. First articulated by Emile Zuckerkandl and Linus Pauling in 1962, the hypothesis emerged from their observation that the number of amino acid differences between hemoglobin sequences from different species scaled roughly linearly with the time since those species diverged in the fossil record. This observation implied that molecular changes accumulate at a steady, predictable rate, effectively acting as a "clock" that ticks at the molecular level. The central claim is not that all genes evolve at the same rate, but that a *particular* gene or protein evolves at a roughly constant rate per unit time across lineages. This property transforms molecular sequences into a powerful tool for estimating divergence times between species, reconstructing phylogenetic relationships, and dating evolutionary events that leave no fossil trace. The [Molecular Clock Definition](/knowledge/molecular-biology/molecular-clock-definition) captures this essential property: a stochastic process in which substitutions accumulate at a rate that is, to a first approximation, proportional to elapsed time. The hypothesis fundamentally changed [evolutionary biology](/blog/guides/evolutionary-biology). Before its formulation, evolutionary timescales were derived almost exclusively from the fossil record, which is incomplete and biased toward hard-bodied organisms. The molecular clock provided an independent, quantitative framework for dating evolutionary divergences, including those among lineages with poor fossil preservation. It also laid the groundwork for the [Neutral Theory of Molecular Evolution](/knowledge/molecular-biology/neutral-theory-of-molecular-evolution), which provides the mechanistic justification for why such a clock might exist in the first place. However, the molecular clock is not a universal constant. The rate of molecular evolution varies among genes, among lineages, and over time. Understanding the mechanistic basis of this variation, the [statistical methods](/blog/guides/statistical-methods) used to model it, and the pitfalls that arise when the clock assumption is violated is essential for any researcher applying these methods. This article provides a mechanistic and methodological treatment of the molecular clock hypothesis, from its theoretical foundations to its practical implementation in modern phylogenetics. ## Mechanistic Basis of the Molecular Clock ### Neutral Theory and Substitution Rates The mechanistic foundation of the molecular clock rests on the [neutral theory of molecular evolution](/knowledge/molecular-biology/neutral-theory-of-molecular-evolution), proposed by Motoo Kimura in 1968. The theory states that the vast majority of molecular changes fixed in a population are selectively neutral—they do not affect organismal fitness. Deleterious mutations are removed by purifying selection, and beneficial mutations are rare and often swept to fixation rapidly, but the bulk of observed substitutions represent the random fixation of neutral or nearly neutral variants through genetic drift. The key insight is that the rate of neutral substitution equals the mutation rate. For a diploid population of effective size \(N_e\), the rate at which new neutral mutations arise is \(2N_e \mu\), where \(\mu\) is the per-generation mutation rate. The probability that any one of these neutral mutations is eventually fixed by drift is \(1/(2N_e)\). The substitution rate \(k\) is therefore: \[ k = 2N_e \mu \times \frac{1}{2N_e} = \mu \] The population size terms cancel exactly. This is the crucial mechanistic result: the rate of neutral substitution is independent of population size and equals the mutation rate per generation. This cancellation explains why a molecular clock can exist despite enormous variation in population sizes across lineages. The per-generation mutation rate \(\mu\) is itself a product of the per-base-pair replication error rate and the number of replication events per generation. In germline cells, DNA replication errors occur at a rate of roughly \(10^{-9}\) to \(10^{-10}\) per base pair per replication, depending on the efficiency of DNA polymerase proofreading and mismatch repair. For a genome of \(3 \times 10^9\) base pairs, this yields on the order of 30–100 new mutations per human generation. The substitution rate is then this mutation rate filtered through the lens of selection: only mutations that are neutral or nearly neutral become fixed at the clock-like rate. The [Neutral Theory of Molecular Evolution](/knowledge/molecular-biology/neutral-theory-of-molecular-evolution) thus provides the mechanistic link between the ticking of the molecular clock and the underlying mutation process. When selection acts on a site, the substitution rate deviates from \(\mu\): purifying selection reduces the rate below \(\mu\), while positive selection can transiently increase it. The clock is therefore most reliable for sites and genes where selective constraints are minimal—synonymous sites in protein-coding genes, non-coding regions, and pseudogenes. ### Generation Time and Metabolic Rate Effects If the substitution rate equals the mutation rate per generation, then the rate per *unit time* depends on the number of generations per unit time. Organisms with shorter generation times—bacteria, rodents, annual plants—experience more rounds of germline replication per year than organisms with long generation times, such as elephants or whales. All else being equal, short-generation organisms should therefore accumulate substitutions faster per calendar year. This generation time effect is well documented. Rodents, for example, evolve at a rate roughly 5–10 times faster than primates for many nuclear genes, reflecting their shorter generation times. The effect is most pronounced in species where the mutation rate is dominated by replication errors rather than by damage-induced lesions repaired by transcription-coupled repair. In mammals, the male germline undergoes more cell divisions per generation than the female germline—spermatogonial stem cells divide continuously throughout life, whereas oocytes are arrested in prophase I—leading to a male-biased mutation rate. This "male-driven evolution" means that substitution rates in mammals correlate with the number of male germline divisions, which in turn correlates with generation time. A second, more controversial effect involves metabolic rate. The hypothesis, advanced by Allan Wilson and colleagues, posits that organisms with higher metabolic rates have higher rates of DNA damage due to increased production of reactive oxygen species (ROS) from mitochondrial oxidative phosphorylation. ROS can cause oxidative lesions such as 8-oxo-guanine, which, if unrepaired, leads to G→T transversions. If repair efficiency is constant, higher metabolic rates should produce more damage and hence more mutations. This effect is most pronounced in [mitochondrial DNA](/blog/guides/mitochondrial-dna), which lacks the protective histone packaging of nuclear DNA and has limited repair capacity. The metabolic rate and generation time effects are not mutually exclusive; both can operate simultaneously. In practice, the relative contribution of each varies among taxa. For example, birds have high metabolic rates but also long generation times, and their nuclear substitution rates are lower than those of mammals, suggesting that generation time dominates in this comparison. In contrast, mitochondrial substitution rates in birds are comparable to or higher than those in mammals, consistent with a metabolic rate effect on mitochondrial genomes. These lineage-specific rate differences are the primary reason why a strict molecular clock—one with a single global rate—is rarely applicable across diverse taxa. The [Molecular Clock Model](/knowledge/molecular-biology/molecular-clock-model) must therefore accommodate rate variation, either by allowing different rates for different lineages (relaxed clocks) or by restricting analyses to closely related species with similar life histories. ## Evidence Supporting the Molecular Clock ### Classic Protein Studies The earliest evidence for the molecular clock came from [protein sequence](/blog/guides/protein-sequence) comparisons. Zuckerkandl and Pauling's original work on hemoglobin showed that the number of amino acid differences between species increased with the time since their divergence, as estimated from the fossil record. This linear relationship held across a wide range of divergence times, from humans versus chimpanzees (approximately 6–8 million years) to humans versus fish (approximately 400–500 million years). Subsequent studies extended these observations to other proteins. The protein cytochrome c, a small heme protein involved in mitochondrial electron transport, was sequenced from a wide range of species. The number of amino acid differences between any two species was found to be roughly proportional to their divergence time, with a rate of approximately 1% amino acid change per 20 million years. Similarly, fibrinopeptides—short peptides cleaved from fibrinogen during blood clotting—evolve rapidly, at roughly 1% per 1–2 million years, because they are under minimal functional constraint. In contrast, histone H4, which is under extreme functional constraint due to its role in [nucleosome assembly](/knowledge/molecular-biology/nucleosome-assembly), evolves at a rate of less than 1% per 500 million years. These observations established a key principle: different proteins tick at different rates, but each protein ticks at a roughly constant rate across lineages. The rate of a protein's clock is inversely related to the strength of purifying selection acting on it, which is itself a function of the protein's functional importance and the fraction of sites that can tolerate substitution. This rate-constraint relationship is a direct prediction of the neutral theory and provides strong support for the mechanistic link between selection, mutation, and the clock. ### DNA Sequence Comparisons With the advent of DNA sequencing, the molecular clock was examined at the nucleotide level. The comparison of synonymous (silent) versus non-synonymous (amino acid-altering) substitution rates provided a powerful test. Synonymous sites, which do not change the encoded amino acid, are under minimal selective constraint and evolve at rates close to the neutral mutation rate. Non-synonymous sites are under purifying selection and evolve more slowly. The ratio of non-synonymous to synonymous substitutions (\(d_N/d_S\)) is therefore a measure of selective constraint, with values near 1 indicating neutrality and values much less than 1 indicating purifying selection. DNA sequence comparisons also revealed that the clock is more reliable over longer timescales. Over short timescales (less than a few million years), stochastic variation in substitution counts can obscure the clock signal. Over very long timescales, multiple substitutions at the same site—saturation—can lead to an underestimate of the true number of substitutions. The [Evidence of Evolution Molecular](/knowledge/molecular-biology/evidence-of-evolution-molecular) from DNA sequences is thus strongest for intermediate timescales, where the number of substitutions is large enough to be statistically meaningful but not so large that saturation becomes a problem. Modern genomic studies have confirmed the broad validity of the molecular clock while also revealing its limitations. Comparisons of orthologous genes across mammals show that synonymous substitution rates are relatively constant across lineages, with a rate of approximately \(10^{-9}\) substitutions per site per year for nuclear genes. Mitochondrial genes evolve faster, at roughly \(10^{-8}\) substitutions per site per year, reflecting the higher mutation rate in mitochondria. These rates are consistent across diverse mammalian orders, supporting the existence of a rough molecular clock, but with notable exceptions—for example, the hominid lineage shows a slower rate than the average mammalian rate, and some rodent lineages show faster rates. ## Statistical Models for Estimating Substitution Rates ### Poisson Process and Gamma-Distributed Rates The simplest statistical model for the molecular clock treats substitutions as a Poisson process. Under this model, the number of substitutions at a site over time \(t\) follows a Poisson distribution with mean \(\lambda t\), where \(\lambda\) is the substitution rate. The probability of observing \(k\) substitutions at a site is: \[ P(k) = \frac{(\lambda t)^k e^{-\lambda t}}{k!} \] This model assumes that all sites evolve at the same rate and that substitutions occur independently. In practice, neither assumption holds. Sites within a gene experience different selective constraints: the active site of an enzyme is more constrained than a surface loop, and a codon's third position is less constrained than its first. Ignoring this rate heterogeneity leads to systematic underestimation of divergence times, because highly constrained sites accumulate few substitutions and dominate the average. To accommodate rate heterogeneity, the gamma distribution is commonly used. The gamma distribution is a flexible two-parameter distribution (shape parameter \(\alpha\) and scale parameter \(\beta\)) that can take a variety of shapes, including highly skewed distributions with many sites evolving slowly and a few evolving rapidly. The shape parameter \(\alpha\) is inversely related to the degree of rate heterogeneity: small \(\alpha\) values (e.g., 0.1–0.5) indicate extreme heterogeneity, while large \(\alpha\) values (e.g., >10) indicate near-uniform rates. The gamma distribution is typically incorporated into substitution models by assuming that the rate at each site is drawn from a gamma distribution with mean 1, and then scaling the overall rate accordingly. The choice of substitution model—whether Jukes-Cantor, Kimura two-parameter, HKY85, or GTR (general time-reversible)—also affects rate estimation. These models differ in their treatment of base frequencies and substitution types. The GTR model, which allows different rates for each of the six substitution types and arbitrary base frequencies, is the most general time-reversible model and is commonly used in modern analyses. Model selection, typically via the Akaike Information Criterion (AIC) or Bayesian Information Criterion (BIC), is essential because using an overly simple model when the true process is complex can bias rate and divergence time estimates. ### Relaxed Molecular Clocks The strict molecular clock assumes a single rate for all lineages. This assumption is frequently violated, as discussed above. Relaxed molecular clock models relax this assumption by allowing rates to vary among lineages, while still providing a framework for estimating divergence times. Two main classes of relaxed clocks exist: uncorrelated and autocorrelated. Uncorrelated relaxed clocks, implemented in programs such as BEAST, assume that the rate for each lineage is drawn independently from a distribution, typically a lognormal or exponential distribution. This model is appropriate when rates are not inherited from ancestor to descendant—for example, when rate changes are driven by lineage-specific factors such as generation time shifts or changes in DNA repair efficiency. Autocorrelated relaxed clocks, in contrast, assume that rates change gradually along branches, with the rate of a descendant lineage being similar to that of its ancestor. This model is appropriate when rates are influenced by slowly evolving factors such as body size or metabolic rate. The choice between uncorrelated and autocorrelated models can be made using model comparison criteria, and the results can differ substantially when rate variation is extreme. The [Molecular Clock Studies](/knowledge/molecular-biology/molecular-clock-studies) literature has shown that relaxed clocks generally provide more accurate divergence time estimates than strict clocks when rate heterogeneity is present, but they also require more data and more careful model specification. The prior distributions on rates and divergence times can strongly influence the results, particularly when the data are limited. ## Methods for Testing Clock-Like Behavior ### Relative Rate Tests Before applying a molecular clock, it is essential to test whether the clock assumption holds. The simplest tests are relative rate tests, which compare the rates of two lineages relative to an outgroup. The logic is straightforward: if two lineages have evolved at the same rate since their divergence from a common ancestor, then the number of substitutions separating each lineage from the outgroup should be equal. The Tajima relative rate test uses a \(2 \times 2\) contingency table of site patterns. For three sequences—two ingroup taxa (1 and 2) and one outgroup (3)—each site is classified into one of four categories: sites where sequences 1 and 2 share a derived state but differ from the outgroup (informative for rate differences), sites where sequence 1 matches the outgroup but sequence 2 differs, sites where sequence 2 matches the outgroup but sequence 1 differs, and sites where all three differ. Under the null hypothesis of equal rates, the number of sites in the second and third categories should be equal. A chi-square test with one degree of freedom is used to assess significance. The relative rate test is simple and requires no model assumptions, but it has low power when the number of substitutions is small. It also assumes that the outgroup is sufficiently distant that shared ancestral polymorphisms do not confound the analysis. More sophisticated relative rate tests, such as the branch-length test, use maximum likelihood to estimate branch lengths under a model that allows different rates for different lineages and compare this to a model with a single rate. ### Likelihood Ratio Tests Likelihood ratio tests (LRTs) provide a more powerful and flexible framework for testing clock-like behavior. The approach compares the likelihood of the data under a model that enforces a strict clock (all tips equidistant from the root) with the likelihood under a model that allows different rates for different branches. The test statistic is twice the difference in log-likelihoods: \[ \Lambda = 2(\ln L_{\text{relaxed}} - \ln L_{\text{strict}}) \] Under the null hypothesis of a strict clock, \(\Lambda\) follows a chi-square distribution with degrees of freedom equal to the difference in the number of parameters between the two models. For a tree with \(n\) tips, the strict clock has \(n-1\) branch lengths (since the tips are equidistant from the root), while the unconstrained model has \(2n-3\) branch lengths, giving \(n-2\) degrees of freedom. The LRT is more powerful than relative rate tests because it uses all the data and accounts for the phylogenetic structure. However, it is sensitive to model misspecification: if the substitution model is incorrect, the test may reject the strict clock even when rates are truly constant. It is therefore essential to select an appropriate substitution model before conducting the LRT. A practical limitation of the LRT is that it tests the strict clock against a fully unconstrained alternative. When the strict clock is rejected, the test does not indicate which lineages deviate from the clock or by how much. In such cases, a relaxed clock model should be used for divergence time estimation, and the specific pattern of rate variation can be examined by comparing branch-specific rates. ## Calibration and Divergence Time Estimation ### Fossil Calibration The molecular clock provides estimates of *relative* divergence times—the ratio of times between different nodes—but converting these to absolute times requires calibration. The most common calibration source is the fossil record, which provides minimum ages for the divergence of particular lineages. For example, the oldest known fossil of a crown-group mammal places a minimum age on the divergence of mammals from other amniotes. Fossil calibrations are typically applied as priors on node ages in Bayesian analyses or as fixed constraints in maximum likelihood analyses. The choice of calibration points is critical. A calibration should be based on a fossil that can be unambiguously assigned to a particular lineage, that preserves enough morphological characters to be placed in a phylogenetic context, and that is old enough to provide a meaningful constraint. Using a misidentified fossil or a fossil that is younger than the true divergence time will bias all downstream estimates. A common error is to treat a fossil age as the *actual* divergence time rather than a *minimum* bound. The fossil record is incomplete; the oldest known fossil of a lineage is almost certainly younger than the lineage's true origin. Calibrations should therefore be specified as minimum bounds, with the prior distribution allowing the true divergence time to be older. In Bayesian analyses, this is typically done using a lognormal or exponential prior with a hard minimum at the fossil age and a soft maximum based on other geological or biological evidence. ### Bayesian Approaches Bayesian methods, implemented in programs such as BEAST and MrBayes, provide a coherent framework for integrating molecular sequence data, fossil calibrations, and models of rate variation. The posterior distribution of divergence times is proportional to the likelihood of the sequence data given the tree and model parameters, multiplied by the prior distribution on those parameters: \[ P(\text{times} | \text{data}) \propto P(\text{data} | \text{times}, \text{model}) \times P(\text{times} | \text{calibrations}) \] Markov chain Monte Carlo (MCMC) is used to sample from the posterior distribution, providing estimates of divergence times with credible intervals that incorporate both sampling error and uncertainty in the calibrations. The choice of priors is a critical and often underappreciated aspect of Bayesian divergence time estimation. The prior on the root age, in particular, can strongly influence the results, especially when the data are not highly informative about deep divergences. Sensitivity analyses, in which the priors are varied and the results compared, are essential for assessing the robustness of the estimates. Bayesian approaches also allow the simultaneous estimation of substitution rates, divergence times, and population sizes, using models such as the coalescent. This is particularly useful for analyses of closely related species or populations, where the assumptions of the molecular clock are more likely to hold and where the focus is on recent demographic events rather than deep divergences. The [Molecular Phylogenetics and Evolution](/knowledge/molecular-biology/molecular-phylogenetics-and-evolution) literature contains numerous examples of such analyses, from dating the origin of HIV subtypes to estimating the timing of human migration out of Africa. ## Common Pitfalls and Misinterpretations ### Overlooking Rate Heterogeneity The most common pitfall in applying the molecular clock is assuming a strict clock when rates vary among lineages. This can lead to systematic errors in divergence time estimates. For example, if a short-generation lineage (e.g., a rodent) is included in an analysis with a strict clock, the high substitution rate in that lineage will cause the clock to tick faster than the true average, leading to underestimation of divergence times for the entire tree. Conversely, excluding fast-evolving lineages can lead to overestimation. Rate heterogeneity among sites within a gene is equally problematic. If the gamma distribution is not used to model among-site rate variation, the estimated divergence times will be biased. The bias arises because highly constrained sites accumulate few substitutions, making the observed number of differences smaller than the true number of substitutions. This is particularly severe for deep divergences, where multiple substitutions at the same site are common. A related issue is the failure to account for compositional heterogeneity—differences in base composition among lineages. If one lineage has a higher GC content than another, the substitution process is not homogeneous, and standard models may infer spurious rate differences. Tests for compositional heterogeneity, such as the chi-square test implemented in programs like PAUP*, should be performed before clock analyses. ### Misuse of Calibration Points Calibration errors are a leading cause of inaccurate divergence time estimates. The most common mistakes include: 1. **Using a single calibration point**: A single calibration provides no cross-validation. If the calibration is wrong, all estimates are wrong. Multiple calibrations, ideally from different time periods and different parts of the tree, are essential. 2. **Treating minimum ages as exact ages**: As noted above, fossil ages are minimum bounds. Treating them as exact ages biases estimates toward younger divergence times. 3. **Using secondary calibrations**: Secondary calibrations are divergence times estimated from other molecular analyses. Using them as calibrations in a new analysis introduces circularity and propagates errors. Primary calibrations from the fossil record are always preferable. 4. **Ignoring the prior distribution**: In Bayesian analyses, the prior on calibration nodes can dominate the posterior if the sequence data are uninformative. Checking the effective sample size (ESS) of the posterior and comparing the prior and posterior distributions for calibration nodes is essential. Another pitfall is the misuse of the molecular clock for dating events that are not divergences. The clock dates the time since the most recent common ancestor of the sampled sequences, which may be much older than the divergence of the species if ancestral polymorphism persists. For closely related species, the coalescent time can be substantially older than the speciation time, and the difference must be accounted for using models that incorporate both speciation and coalescence. ## Practical Summary and Best Practices The molecular clock hypothesis remains a cornerstone of molecular evolution, but its application requires care. The following best practices will help avoid the most common errors: 1. **Test the clock assumption**: Use relative rate tests and likelihood ratio tests to determine whether a strict clock is appropriate. If the strict clock is rejected, use a relaxed clock model. 2. **Model rate heterogeneity**: Always incorporate among-site rate variation using a gamma distribution or an equivalent model. Select the substitution model using formal criteria such as AIC or BIC. 3. **Use multiple, well-vetted calibrations**: Choose fossils that are unambiguously assigned to lineages, treat their ages as minimum bounds, and use at least two or three independent calibrations spread across the tree. 4. **Perform sensitivity analyses**: Vary the priors, the calibration points, and the clock model to assess the robustness of the results. If the estimates change substantially, the data may be insufficient to support strong conclusions. 5. **Report uncertainty**: Divergence time estimates should always be reported with confidence or credible intervals, and the sources of uncertainty (sampling error, calibration uncertainty, model uncertainty) should be explicitly discussed. 6. **Be aware of lineage-specific effects**: Generation time, metabolic rate, and DNA repair efficiency can all affect substitution rates. When comparing distantly related lineages, consider whether these factors are likely to be similar. ## Frequently Asked Questions ### What is the molecular clock hypothesis? The molecular clock hypothesis is the proposition that nucleotide or amino acid substitutions accumulate at a roughly constant rate over time for a given gene or protein. This constancy allows molecular sequences to be used as a "clock" for estimating the time since two lineages diverged from a common ancestor. ### How does the molecular clock work? The clock works because the rate of neutral substitution equals the mutation rate, and the mutation rate per generation is roughly constant for a given species. By measuring the number of differences between two sequences and dividing by the substitution rate, one can estimate the time since their divergence. ### What is the [neutral theory of molecular evolution](/knowledge/molecular-biology/neutral-theory-of-molecular-evolution)? The neutral theory, proposed by Motoo Kimura, states that most molecular changes fixed in a population are selectively neutral and become fixed through genetic drift rather than natural selection. The theory provides the mechanistic basis for the molecular clock by showing that the substitution rate equals the mutation rate for neutral sites. ### What are the limitations of the molecular clock? The clock is limited by rate variation among lineages (due to generation time, metabolic rate, and other factors), rate variation among sites within a gene, saturation of substitutions at deep divergences, and the difficulty of calibrating the clock using an incomplete fossil record. ### How is the molecular clock calibrated? Calibration involves using external information—typically fossil ages, but also biogeographic events such as island formation or the rise of the Isthmus of Panama—to convert relative divergence times into absolute times. Calibrations are applied as priors on node ages in Bayesian analyses or as fixed constraints in likelihood analyses. ### What is a relaxed molecular clock? A relaxed molecular clock is a model that allows substitution rates to vary among lineages, rather than assuming a single global rate. Uncorrelated relaxed clocks assume rates are drawn independently from a distribution, while autocorrelated relaxed clocks assume rates change gradually along branches. ### What are relative rate tests? Relative rate tests compare the number of substitutions in two lineages relative to an outgroup to determine whether the two lineages have evolved at the same rate. The Tajima test uses a contingency table of site patterns and a chi-square test to assess significance. ## Key Takeaways - The molecular clock hypothesis states that substitutions accumulate at a roughly constant rate per unit time for a given gene, enabling divergence time estimation from sequence data. - The neutral theory provides the mechanistic basis: the neutral substitution rate equals the mutation rate, independent of population size. - Generation time and metabolic rate cause lineage-specific rate variation, making strict clocks inapplicable across diverse taxa. - Statistical models must account for among-site rate heterogeneity, typically using a gamma distribution, and for among-lineage rate variation using relaxed clocks. - Fossil calibrations are minimum bounds, not exact ages; multiple, well-vetted calibrations are essential for accurate divergence time estimates. - Testing the clock assumption with relative rate tests and likelihood ratio tests is a prerequisite for any clock-based analysis. - Common pitfalls include overlooking rate heterogeneity, misusing calibration points, and ignoring lineage-specific effects; sensitivity analyses are essential for robust conclusions. ## Further Reading - dos Reis M, Donoghue PC, Yang Z. *Bayesian molecular clock dating of species divergences in the genomics era*. Nature reviews. Genetics. 2016. [PubMed 26688196](https://doi.org/10.1038/nrg.2015.8) - Li WH. *So, what about the molecular clock hypothesis?*. Current opinion in genetics & development. 1993. [PubMed 8118215](https://doi.org/10.1016/0959-437x(93)90011-d) - Kumar S. *Molecular clocks: four decades of evolution*. Nature reviews. Genetics. 2005. [PubMed 16136655](https://doi.org/10.1038/nrg1659) - Easteal S. *A mammalian molecular clock?*. BioEssays : news and reviews in molecular, cellular and [developmental biology](/blog/careers/developmental-biology). 1992. [PubMed 1503557](https://doi.org/10.1002/bies.950140613) - Hu T et al. *The genetic equidistance result: misreading by the molecular clock and neutral theory and reinterpretation nearly half of a century later*. Science China. Life sciences. 2013. [PubMed 23526392](https://doi.org/10.1007/s11427-013-4452-x) - Behe MJ. *Histone deletion mutants challenge the molecular clock hypothesis*. Trends in biochemical sciences. 1990. [PubMed 2251727](https://doi.org/10.1016/0968-0004(90)90231-y)
## Related Clinical & Scientific Guides * [MAPK Pathway: Mechanism, Function, and Clinical Relevance](/knowledge/molecular-biology/mapk-pathway) * [Mammalian Cell Culture Bioreactors: A Practical Guide](/knowledge/molecular-biology/mammalian-cell-culture-bioreactor) * [Nucleotide Formation: Biosynthesis and Assembly of DNA/RNA Building Blocks](/knowledge/molecular-biology/nucleotide-formation)