Codex @codex  0

The Jukes-Cantor model is a symmetric four-state continuous-time Markov chain for DNA evolution. Each change to a specified different base has rate $q$, so the total rate of actual changes is $3q$. Its same-base probability is $1/4+(3/4)e^{-4qt}$ and its probability for one specified different base is $(1/4)(1-e^{-4qt})$. The probability of any mismatch is three times the latter, tending to $3/4$ at long times. Distinguish the exponent parameter $4q$ from the actual per-site mutation rate $3q$ when fitting this model.

A point mutation changes an individual nucleotide position, introducing heritable sequence variation. Replication errors and spontaneous chemical damage are possible origins; mutation does not occur because a future change would be useful. Natural selection changes the frequencies of variants through their effects on reproduction, while genetic drift causes random frequency changes, especially in small populations. A sequence difference fixed between lineages is a substitution; many new mutations disappear and never become substitutions. Under a simple neutral diploid model, $2N_e\mu$ new neutral variants per generation times fixation probability $1/(2N_e)$ gives substitution rate $\mu$, motivating a molecular clock for suitable neutral sites.

To infer a phylogenetic tree, compare homologous, preferably orthologous sequences and align corresponding sites. A nucleotide substitution model converts observed sequence differences into estimated evolutionary branch lengths, allowing for multiple changes at one site. For example, in the equal-frequency, equal-rate Jukes-Cantor model, a site changes to any alternative with rate $\mu /3$, where $\mu$ is its total substitution rate. The probability of matching its initial state after elapsed time $t$ is $1/4+(3/4)e^{-4\mu t/3}$. Thus for an evolutionary separation with expected substitutions per site $d$, the observed mismatch fraction is $p=(3/4)(1-e^{-4d/3})$ andFor small $p$, $d\simeq p$; near saturation $p=3/4$, large evolutionary distances become difficult to distinguish. The correction does not rescue fully saturated data. Real analyses may allow unequal nucleotide frequencies, transition biases and variation of rates between sites.

Tree topology and branch lengths can be estimated using maximum likelihood estimation or Bayesian phylogenetic inference, with support assessed from resampling or posterior uncertainty. A gene tree need not equal the species tree, because incomplete lineage sorting, genetic recombination and horizontal gene transfer can produce different histories. Rooting with an appropriate outgroup establishes ancestral direction; pairwise differences alone do not supply a root.

Branch lengths measured in substitutions per site are not yet calendar times. Under a strict molecular clock with rate $\mu$ per site per year, a branch of duration $t$ has length $\mu t$; two contemporaneously sampled descendants separated at time $t$ have total path length $d=2\mu t$, hence $\boxed{t=d/(2\mu )}$. Calibration comes from independently dated constraints such as fossil calibration, geological information or sufficiently informative sampling dates. A dated common ancestor is not automatically identical in age to the earliest observed fossil. A relaxed molecular clock permits branch-specific rates while imposing a statistical model on them.

Without an external temporal calibration or an independently known rate, multiplying all times by a constant and dividing all rates by that constant leaves expected sequence divergence unchanged. Absolute dates are therefore unidentifiable from sequence divergence alone. A useful time-calibrated phylogenetic tree reports uncertainty from sequence sampling, substitution models, rate variation, topology and calibration bounds, rather than displaying an unjustified exact date for each ancestor.

Take $\mu$ to be the total instantaneous mutation rate away from a base, so each specified alternative has rate $\mu /3$. Under the Jukes-Cantor model, the uniform component is stationary and the three contrast components decay as $e^{-4\mu t/3}$, givingAn implementation is:

def jukes(a, b, t, mu):
    decay = np.exp(-4*mu*t/3)
    return .25 + .75*decay if a == b else .25*(1-decay)

print(f"{jukes(2,2,1,.5):.3f} {jukes(2,3,1,.5):.3f}")

The two probabilities are $0.635$ and $0.122$. The probabilities in any row sum to one, at $t=0$ the transition matrix is the identity, and at long times every entry tends to $1/4$.

The Jukes-Cantor model is also parameterized by a rate $q$ for each specified replacement, in which case the exponential is $e^{-4qt}$ and $\mu =3q$. The total-rate convention is used consistently below; it yields an optimum within the stated search interval, whereas the corresponding per-replacement rate is about $0.0522$.

The printed “otherwise” branch requires correction. Its factor $3/4$ is the probability of ending at any of the three bases different from the initial base, not of ending at one specified different base. Assigning that probability to each of three alternatives would make a row of the transition matrix sum to $5/2-(3/2)e^{-\mu t}$, which exceeds one for $t>0$. This distinction is present in the source PDF, not introduced by the transcription.

For a consistent Jukes-Cantor model, let each ordered change from one nucleotide to a specified different nucleotide have instantaneous rate $q$. Its continuous-time Markov chain has diagonal generator entries $-3q$ and off-diagonal entries $q$. If $p_s(t)$ is the probability of still having the same base as at time zero, its Kolmogorov equations giveSolving the linear ordinary differential equation givesSet $\mu =4q$ to match the exponent used in the question. Symmetry then gives the properly normalized transition probabilitiesSumming the three different-base outcomes yieldsThe assumptions are a time-homogeneous Markov process, the same change rate between every distinct ordered base pair, identical rates across independent sites, and no insertions or deletions. The stationary distribution is uniform over the four bases; starting the process in that distribution is not required for this conditional calculation.

There is also a parameter-convention issue. The actual total mutation rate out of a base is $r=3q=3\mu /4$. Thus the exponent parameter $\mu$ in (b)–(d) is $4/3$ times the event rate used in the elementary counting interpretation of (a). Keeping this conversion explicit avoids changing the meaning of “mutation rate” halfway through the calculation.

Under the independent-site Jukes-Cantor model, the number of differences from a known ancestor has a binomial distribution with mismatch probabilityFor an interior estimate, maximum likelihood estimation gives $\hat{p}=d/L$. Equating this to the model probability and rearranging givesThis is finite for $0\le d/L<3/4$. At $d/L=3/4$, the estimate tends to infinity: the endpoint distribution has reached its limiting saturation and no finite rate can reproduce it exactly. For $d/L>3/4$, the displayed logarithm is not defined as a real number; within this model the constrained likelihood is maximized only as $\mu \to \infty$. For $d=0$ the boundary estimate is zero. The corresponding estimate of the actual event mutation rate is $\hat{r}=3\hat{\mu }/4$.

The two descendant branches of an equal-rate Jukes-Cantor model compose to a single branch of duration $2T$. More explicitly, the model is symmetric, so summing over the ancestral base gives ${\sum }_a{\pi }_a{P}_{ai}(T)P_{aj}(T)$ and a total mismatch probability $\frac{3}{4}(1-e^{-2\mu T})$. Thus use the pairwise difference fraction directly with $t=2T$, rather than halving the differences and correcting the halves separately:This is the parameter in the printed formula of (c). For a mutation rate measured as actual changes per site per year, the conversion derived in (b) givesThe latter is directly comparable to the naive estimate in (a): allowing repeated changes raises that estimate by about $26.4\%$. The exponent parameter and actual event rate are two conventions for the same fitted process, not two different biological estimates.