A 2025 bioRxiv preprint by Saini, Usmanova, and Vitkup resolved the long-standing absence of a detectable generation time effect for amino-acid substitutions in vertebrates: across 43 pairs of mammalian sister species, per-year substitution rates scale strongly with generation time, and an integrated model that traces generation-time changes along the mammalian phylogeny reproduces protein sequence divergence over roughly 160 million years of evolution. The preprint itself frames what remains open: whether this framework generalizes beyond mammals to vertebrate and invertebrate species with broader life histories and on longer timescales; whether life-history data from the fossil record (for example body mass) can anchor estimates of ancestral generation times; and whether the between-species life-history effects can be merged with within-species structural constraints and functional optimization into a single quantitative theory of protein evolution. The paper additionally acknowledges two model limitations that it does not frame as future directions: the assumption that generation time changes linearly along branches, and the fact that linked selection (background selection and genetic hitchhiking) was not modeled. None of these extensions has yet been undertaken, so the problem is open.

Background

The protein molecular clock is the observation, foundational to molecular evolution, that each protein accumulates amino-acid substitutions at an approximately constant, protein-specific rate over long evolutionary timescales. Under the neutral theory of molecular evolution, most fixed substitutions have negligible effects on fitness, so the substitution rate equals the mutation rate; because new mutations arise mainly during germ-line cell division, species with short generation times should evolve faster per unit of real time than species with long generation times. This predicted 'generation time effect' was indeed observed for synonymous substitutions (mutations that do not change the encoded amino acid), but for decades it could not be detected for non-synonymous (amino-acid-changing) substitutions in vertebrates, calling a core tenet of the neutral theory into question. Proposed resolutions included the nearly neutral theory — slightly deleterious substitutions fix less often in species with large effective population size (the number of individuals in an idealized population with the same strength of genetic drift as the real one), and short-generation species tend to have large populations, which can offset their higher mutation input — and episodic adaptive bursts timed to real-world environmental change rather than to generations.

In a January 2025 preprint, Saini, Usmanova, and Vitkup attacked this question with modern comparative-genomic resources: 43 pairs of recently diverged mammalian sister species, 2,578 one-to-one orthologous genes (genes in different species that descended from a single gene in their last common ancestor, with one copy in each species), direct mutation-rate estimates from parent-offspring trio sequencing (whole-genome sequencing of parents and their offspring to observe newly arising germline mutations directly) in 23 species, and curated generation times. They found a pronounced generation time effect for both non-synonymous and synonymous per-year substitution rates (for example, a correlation of r=−0.69r=-0.69 between the per-year non-synonymous rate dNd_N and generation time). Two measured mechanisms partially offset, but do not eliminate, the effect: mutation-rate selection (long-generation species evolve higher per-generation mutation rates, log-log slope ≈0.26\approx 0.26) and stronger purifying selection in short-generation species (captured by the ratio dN/dSd_N/d_S of non-synonymous to synonymous substitution rates, slope ≈0.22\approx 0.22). With both mechanisms included, the observed scaling is fully consistent with predominantly neutral substitutions. They then built an integrated model of the clock: at each internal node of the phylogeny the ancestral generation time is estimated as the harmonic mean of the two daughter species' generation times, and each lineage's effective generation time is assumed to change linearly along its branch from the ancestral to the present value. Accumulating substitutions along the tree this way, the model reproduces the observed approximately linear growth of sequence divergence with time for 323 proteins across human and 177 other mammals over roughly 160 million years (r=0.95r=0.95), reconciling the generation time effect with the time-like behavior of the protein clock.

The preprint's discussion is explicit that the framework was constructed and validated only on mammals and on extant-species generation times, and it identifies the directions left open: extension to vertebrate and invertebrate species with broader life histories and longer timescales; use of life-history information from the fossil record (the works it cites concern body-mass evolution in fossil mammals) to estimate ancestral generation times; and combination of these life-history effects with structural constraints and functional optimization — the same group's earlier work (Usmanova, Plata, and Vitkup, 2024) showed that selection to minimize expression costs, for example by tuning enzymatic kinetic constants in specific tissues, slows the evolution of highly expressed proteins — into an integrated functional theory of protein evolution describing rates both within and between species. Related limitations acknowledged in the paper are the model's simplifying assumption that generation time changes linearly along branches, and its neglect of linked selection, i.e. background selection (the loss of neutral variants located near deleterious selected sites) and genetic hitchhiking (the spread of neutral variants linked to advantageous ones), whose influence on the observed generation-time scaling was not modeled.

Problem Statement

The problem, stated as one question with a single scientific objective — to establish a general, validated, integrated quantitative account of how life history and functional constraints jointly determine protein substitution rates — is to extend and complete the integrated life-history model of the protein molecular clock beyond its current mammalian formulation based on extant-species generation times. Concretely: (1) test the integrated model across both vertebrate and invertebrate species, whose broader range of life histories and deeper evolutionary timescales probe whether the mammalian generation-time scaling and its mutation-rate-selection and functional-selection offsets are general; (2) integrate life-history data derived from the fossil record (for example body mass) to improve the estimation of ancestral generation times and to test the model on longer evolutionary timescales; and (3) combine the between-species life-history effects with within-species structural constraints and functional optimization (such as variability in enzymatic kinetic constants and expression costs) into a single integrated functional theory of protein evolution capable of describing protein evolution both within and between species. Components (1) and (2) test the model's generality across taxa and timescales, and component (3) completes it into a unified theory; together they decide whether the integrated account is a general theory of protein evolution or a mammalian-specific description. The account is general insofar as the integrated model reproduces the observed divergence-versus-time relationship across both vertebrate and invertebrate taxa beyond the mammalian species of the source model, ancestral generation times reconstructed from fossil-derived life-history data (for example body mass) extend that relationship to an evolutionary timescale deeper than the roughly 160-million-year mammalian window, and a single account reproduces the observed within-species and between-species patterns of rate variation. Whether the mammalian generation-time scaling extends to a given clade is itself an open scientific question of the same scope.

References

  1. Saini A, Usmanova DR, Vitkup D. Integrated model of the protein molecular clock across mammalian species. bioRxiv preprint posted 2025-01-15. DOI: 10.1101/2025.01.14.633008. URL: https://doi.org/10.1101/2025.01.14.633008
  2. Saini A, Usmanova DR, Escalante RS, Vitkup D. Multiple molecular and cellular properties jointly affect protein and site-specific evolutionary rates. bioRxiv preprint posted 2026-05-23. DOI: 10.64898/2026.05.20.726710. URL: https://doi.org/10.64898/2026.05.20.726710
  3. Herrick J. DNA Replication Timing, Genome Stability and Non-adaptive Radiation. arXiv preprint arXiv:2511.12183, 2025. DOI: 10.48550/arXiv.2511.12183. URL: https://arxiv.org/abs/2511.12183
  4. Sterling JE, Zwonitzer KD, Havird JC. Lifespan Predicts Mitochondrial Substitution Rates across Vertebrates, but Methodology Matters. Genome Biology and Evolution 18(3):evag067, 2026. DOI: 10.1093/gbe/evag067. URL: https://doi.org/10.1093/gbe/evag067
  5. Cai T, Wen Z, Jiang Z, Zhen Y. Distinct latitudinal patterns of molecular rates across vertebrates. Proceedings of the National Academy of Sciences 122(19):e2423386122, 2025. DOI: 10.1073/pnas.2423386122. URL: https://doi.org/10.1073/pnas.2423386122
  6. Usmanova DR, Plata G, Vitkup D. Functional Optimization in Distinct Tissues and Conditions Constrains the Rate of Protein Evolution. Molecular Biology and Evolution 41(10):msae200, 2024. DOI: 10.1093/molbev/msae200. URL: https://doi.org/10.1093/molbev/msae200