Evolutionary trees from DNA sequences: A maximum likelihood approach¶
Why this mattered¶
Felsenstein’s 1981 paper shifted molecular phylogenetics from primarily distance-based or parsimony-based tree reconstruction toward an explicitly statistical framework. Its central advance was to define the probability of observed DNA sequences given a tree, branch lengths, and a model of nucleotide substitution, then choose the tree that maximizes that likelihood. This made evolutionary trees estimable under stated assumptions about sequence change rather than only by minimizing changes or clustering pairwise distances.
The paper also made maximum-likelihood phylogenetics computationally practical through the recursive likelihood calculation now widely known as Felsenstein’s pruning algorithm. That algorithm allowed likelihoods to be evaluated efficiently by summing over unobserved ancestral states without enumerating every possible ancestral sequence assignment. After this, researchers could compare tree topologies, estimate branch lengths, and later test evolutionary models within one probabilistic framework.
Its importance reaches well beyond the original DNA substitution setting. The same likelihood logic became the foundation for model-based phylogenetics, including likelihood-ratio tests, codon and rate-variation models, molecular dating, and Bayesian phylogenetic methods developed in the 1990s and 2000s. In effect, the paper helped turn phylogeny inference into a modern statistical discipline, enabling later breakthroughs that depended on explicit evolutionary models and computable probabilities over trees.
Abstract¶
(no abstract available)
Related¶
- cite → Maximum Likelihood from Incomplete Data Via the EM Algorithm — Felsenstein’s phylogenetic likelihood method cites EM as a general maximum-likelihood strategy for models with unobserved latent variables.
- cite ← Phylogenies and the Comparative Method — Felsenstein's comparative method builds on maximum-likelihood phylogenetic tree estimation for modeling trait correlations along evolutionary trees.