Adaptive molecular convergence persists across medusozoan lineages separated by more than 600 million years
A new PNAS analysis argues that adaptive molecular convergence is far more common across deep evolutionary time than many comparative studies have assumed. In Medusozoa, repeated protein-level convergence did not vanish as lineages diverged. It lingered across species separated by more than 600 million years, where excess convergence in nonsynonymous substitutions continued to exceed null expectations based on random overlap.
The finding cuts against a familiar habit in evolutionary genomics: identify a strikingly similar trait, then look for the same genes changing in parallel. Here, the molecular pattern did not simply track the visible phenotype. Eyes, medusa loss, and upright colonies all appeared repeatedly in this animal group, yet those repeated traits did not correspond to higher molecular convergence than other lineage comparisons. The genomic signal was more scattered, more contingent, and far less phenotype-bound than the trait-centric search strategy would predict.
Excess protein-level convergence was not confined to closely related species
The analysis combined phylotranscriptomics with sequence-evolution methods to estimate convergence across the medusozoan tree. Adaptive molecular convergence was defined as excess convergence of nonsynonymous substitutions between homologs, a pattern consistent with positive selection rather than neutral coincidence. Across the dataset, that signal was not rare or isolated. It remained detectable even as evolutionary distance increased, declining slightly over time but never collapsing into background noise.
A single sentence captures the central result: similar protein changes kept appearing in lineages that had been apart for immense spans of time.
For comparative biology, this matters because deep-time convergence is often treated as fragile evidence, easily eroded by substitutional saturation and statistical noise. The present analysis points in the opposite direction. The authors’ framework suggests that selection can repeatedly arrive at similar amino acid states across very distant branches, even when the outward trait looks nothing like a simple molecular repeat. In other words, the protein sequence can converge without the phenotype announcing a clean molecular script.
Shared phenotypes did not predict stronger genomic reuse
The study also tested a straightforward expectation: lineages that independently evolved the same phenotypes should show stronger reuse of the same genes. That pattern did not emerge. Comparisons among species that shared recurrent traits such as eyes, medusa loss, and upright colonies did not display elevated adaptive molecular convergence relative to other pairs.
This is a sharp result because it separates two ideas that are often linked too loosely. Phenotypic convergence describes what organisms look like or how they function in the world. Molecular convergence describes how their proteins change under selection. The two can overlap, but they do not have to. In this dataset, they were broadly decoupled.
The implication is not that phenotype is irrelevant. It is that the same trait can arise through different molecular routes, and the same molecular route can appear in lineages with no obvious shared morphology. Evolutionary outcomes appear to be shaped by constraints, contingencies, and ecological pressures that operate below the level of visible form.
Environmental genes carried most of the signal
The convergent substitutions were not randomly distributed across the genome. They were concentrated in genes tied to environment-facing functions, including metabolism, immunity, and xenobiotic processing. Those categories fit the biology of organisms constantly negotiating chemical, microbial, and nutritional conditions in their habitats. They also make mechanistic sense: selection on proteins that interface directly with the environment is likely to recur when organisms confront similar external challenges, even if those challenges are embedded in different life histories.
That concentration helps explain why the signal survived across deep time. Genes involved in environmental interaction are frequently exposed to changing selective pressures, and similar biochemical problems can recur across lineages that occupy comparable ecological settings. The molecular changes are therefore better understood as responses to repeated functional demands than as a simple byproduct of shared ancestry or shared body plans.
The result also argues against a narrow view of convergence as a feature of famous morphological case studies alone. Adaptive convergence appears to be a broader genomic phenomenon, especially in pathways that mediate contact between organism and environment. Eyes and colonial architecture may grab attention, but the stronger evolutionary logic may sit in metabolic tuning, immune defense, and detoxification chemistry.
Selection seems to work through lineage-specific ecological histories
The strongest interpretive thread in the paper is that molecular convergence is real, but its causes are not unified. Similar substitutions can arise in unrelated lineages, yet the selective pressures producing them appear to be comprehensive and lineage-specific. Rather than a single recurring evolutionary template, the data point to many local solutions to many local problems.
This is an important corrective to overly tidy stories about repeatability. Adaptive evolution can be directional without being uniform. It can produce similar amino acid changes without producing the same phenotype in every lineage, and it can produce similar phenotypes through distinct molecular routes. The study places more weight on ecological context than on a simple one-gene, one-trait framework.
For evolutionary genomics, the practical consequence is methodological as much as conceptual. Trait-first scans for convergence may miss the broader genomic architecture of selection. If molecular convergence is widespread but only loosely coupled to obvious phenotype convergence, then analyses anchored exclusively to conspicuous traits will systematically undercount how often selection reuses protein sequence space.
Deep-time convergence may be common, but it is not uniform
The persistence of adaptive molecular convergence across more than 600 million years is striking, yet the pattern was not even across all comparisons. Convergence occurred idiosyncratically among species pairs, rather than as a universal property of every line of descent. That unevenness matters. It suggests that repeated selection can be powerful without being predictable in a simple pairwise sense.
The work leaves evolutionary biologists with a harder but more realistic picture of repeatability. Molecular evolution is not a neat replay of the same script whenever a similar phenotype appears. It is a dispersed process, shaped by environment-facing functions, historical contingencies, and the particulars of each lineage’s ecological exposure. The study also raises a technical caution for future work: deep-time convergence can be real even when it does not align cleanly with the phenotypes most often used to frame comparative questions.
What this analysis does not resolve is the full causal chain behind each convergent substitution. The paper establishes that selection often drives similar protein changes across disparate lineages, but it does not collapse those changes into a single ecological explanation. The challenge now is to connect convergent amino acid substitutions to the specific environmental pressures that selected them, one lineage at a time.