Comparison of the Distribution of Fitness Effects Across Primates

By analyzing the distribution of fitness effects across 38 catarrhine primates, this study demonstrates that interspecific differences in deleterious mutations are primarily driven by variation in effective population size rather than clade-specific factors, supporting the nearly neutral theory and confirming the robustness of these findings to demographic history and dominance effects.

The Gist

The Big Picture: The "Fitness Landscape" of Primates

Imagine the genome of a primate (like a human, a chimp, or a monkey) as a massive, complex video game world. In this world, every time a player (a cell) makes a move (a mutation), it changes the character's stats. Some moves make the character super strong (beneficial), some make them slightly clumsy (mildly harmful), and some are instant game-overs (lethal).

The Distribution of Fitness Effects (DFE) is essentially a report card that tells us: "Out of all the possible random moves a player could make, what percentage are good, bad, or deadly?"

This paper asks a huge question: Is this report card the same for all primates, or does it change depending on the species?

The Main Discovery: It's Not the Game, It's the Crowd Size

The researchers looked at 38 different primate species (from great apes to various monkeys). They found something fascinating: The "rules" of the game (the underlying fitness landscape) are almost exactly the same for everyone.

However, the experience of playing the game feels very different depending on how many people are playing at once. This is the key concept of Effective Population Size ($N_e$).

• The Analogy: Imagine two groups of people trying to fix a broken car.
  • Group A (Small Population): A tiny village with only 10 people. If someone suggests a bad fix, the group might accidentally keep it because they don't have enough people to argue against it. They are easily swayed by "drift" (random chance).
  • Group B (Large Population): A massive city with 10,000 mechanics. If someone suggests a bad fix, the sheer number of people ensures it gets spotted and rejected immediately. They have high "selection efficiency."

The Paper's Finding:
The study found that species with larger populations (like some macaques) are like the massive city. They are so good at spotting and removing bad mutations that their genetic "report card" looks like they have a higher percentage of "strongly bad" mutations. Why? Because they are so efficient at purging the "mildly bad" ones that the only ones left are the truly terrible ones.

Conversely, species with smaller populations (like humans or chimps) are like the tiny village. They keep a lot of "mildly bad" mutations because they aren't efficient enough to weed them all out.

The Takeaway: The difference in the "report cards" isn't because humans and monkeys have different biological rules; it's just because the "village" (humans) is smaller than the "city" (some monkeys), so the village keeps more junk.

The "Hidden" Factor: Dominance (The Masking Effect)

The researchers also asked: "What if bad mutations are 'recessive'?"

• The Analogy: Imagine a bad mutation is a broken lightbulb.
  • Additive (No Masking): If you have one broken bulb, the light is 50% dim. Everyone sees it.
  • Recessive (Masking): If you have one broken bulb and one good one, the good one covers for it, and the light looks normal. The problem is "hidden" until you have two broken bulbs.

The paper tested if assuming mutations are "hidden" (recessive) changes the results. They found that while it does shift the numbers slightly, it doesn't change the main story. The "masking" effect is real, but it doesn't explain the differences between species. The main driver is still the population size.

The "Beneficial" Mutations: The Rare Gold Coins

The study also looked at beneficial mutations (the "good moves").

• The Finding: Species with larger populations tend to find and keep more beneficial mutations.
• The Analogy: In a city of 10,000 people, if a genius invents a new tool, it's likely to be noticed and adopted. In a village of 10, if a genius invents the same tool, they might just die of old age before anyone notices.
• The Caveat: The researchers were careful to say that finding these "good" mutations in the data is very hard. It's like trying to find a single gold coin in a pile of sand; the signal is very weak.

Why This Matters

1. Conservation: It suggests that the fundamental "rules of life" (how mutations affect fitness) haven't changed much in primates over the last 30 million years. We are all playing by the same rulebook.
2. Demography is King: When we see differences in genetic health between species, we shouldn't immediately blame "different biology." We should first look at population size. A small population is naturally "sloppier" at cleaning up genetic errors.
3. Methodology: The paper proves that their computer models (using something called "Site Frequency Spectra") are robust. Even when you account for complex family trees, population crashes, or hidden recessive traits, the main conclusion holds up.

In a Nutshell

Think of evolution as a quality control process.

• Large populations are like strict, high-volume factories. They catch almost every tiny defect, so the final product looks very "pure," but the process is intense.
• Small populations are like small, artisanal workshops. They let a few minor defects slip through because they don't have enough inspectors.

This paper confirms that the blueprint (the genome's potential) is the same for all primates. The quality control (how well they clean up mistakes) is what changes, and that depends entirely on how big the workshop is.

Technical Summary

1. Problem Statement

The Distribution of Fitness Effects (DFE) of new mutations is a fundamental parameter in evolutionary biology, determining the efficacy of natural selection, genetic load, and the rate of adaptation. While the DFE provides a compressed summary of the underlying fitness landscape, comparing DFEs across species presents a significant conceptual and methodological challenge:

• Unit Interchangeability: Experimental evolution measures unscaled selection coefficients ($s$), whereas population genomic inference based on the Site Frequency Spectrum (SFS) typically yields population-scaled coefficients ($S = 4N_e s$).
• Confounding Factors: Differences in $N_e$ (effective population size) can mimic differences in the underlying fitness landscape ($s$). If not properly disentangled, observed variations in DFEs may reflect differences in genetic drift regimes rather than true changes in the fitness landscape topology.
• Dominance Uncertainty: Most deleterious mutations are partially recessive. However, dominance coefficients ($h$) are notoriously difficult to infer from SFS data, and standard comparative studies often assume additive effects ($h=0.5$), potentially biasing cross-species comparisons.

The authors aim to resolve these issues by inferring DFEs across 38 catarrhine primate subspecies to determine if the fitness landscape is conserved or if observed differences are driven primarily by variation in $N_e$.

2. Methodology

Data and Preprocessing

• Dataset: A subset of 38 catarrhine subspecies (1,873 individuals) from the Primate Diversity Panel (Bergman et al., 2025), covering 11 great apes, 14 macaques, 6 baboons, and other Old World monkeys.
• Reference Genomes: To avoid coordinate transfer biases in distantly related species, the authors used species-specific reference genomes rather than a single common reference.
• Ancestral State Inference: Custom workflows using Minimap2 for pairwise whole-genome alignments were developed to lift outgroup sequences onto ingroup coordinates. Ancestral alleles were inferred using the EST-SFS model (Keightley and Jackson, 2018) via fastDFE.
• Site Classification: Coding sites were classified as 4-fold degenerate (neutral) or 0-fold degenerate (selected) to generate separate SFS inputs.

Inference Framework (fastDFE)

The study utilized the fastDFE software package with two primary parameterizations:

1. Gamma-Exponential Model: Deleterious effects follow a reflected gamma distribution; beneficial effects follow an exponential distribution.
2. Discrete Model: Assigns weights to predefined intervals of $S$ (e.g., $[-\infty, -100], [-100, -10], \dots$) to avoid parametric assumptions.

Key Extensions:

• Dominance Modeling: The authors extended fastDFE to jointly infer or fix dominance coefficients ($h$). They implemented a model where $h$ covaries with selection strength ($S$), specifically testing a function where weakly selected mutations are nearly additive and strongly deleterious ones are recessive ($h(S) = 0.4e^{-0.02|S|}$).
• Demographic Correction: Nuisance parameters derived from the neutral SFS were used to correct for non-equilibrium demographic histories (bottlenecks, expansions).

Statistical Analysis

• Regression: Associations between DFE properties and $\log_{10}(N_e)$ were tested using Ordinary Least Squares (OLS) and Phylogenetic Generalized Least Squares (PGLS) with Brownian motion and Grafen's branch-length transformation.
• Simulations: Extensive forward simulations using SLiM were performed to validate inference accuracy under various demographic scenarios, dominance models, and sample sizes.

3. Key Contributions

1. Resolution of the $N_e$ vs. $s$ Conundrum: The study rigorously demonstrates that while DFEs appear distinct when viewed in scaled units ($S$), they are largely conserved in unscaled units ($s$) across primates. This suggests the underlying fitness landscape is invariant, and observed differences are driven by variation in $N_e$.
2. Robust Dominance Modeling: The authors provide a framework for incorporating non-additive dominance into DFE inference. They show that while $h$ is weakly identifiable from SFS alone, assuming partial recessivity introduces only a modest, systematic bias that does not alter cross-species comparative conclusions.
3. Methodological Validation: The paper validates fastDFE against complex demographic scenarios (bottlenecks, structure) and demonstrates that demographic correction via nuisance parameters enables robust inference even when the population is not at equilibrium.
4. Re-evaluation of Previous Findings: The study re-analyzes data from Castellano et al. (2019) regarding great apes, finding that previously reported outliers (bonobos and western chimpanzees) and specific correlations between $N_e$ and unscaled $s$ were likely artifacts of parametrization or statistical noise, not robust biological signals.

4. Key Results

• Conservation of the Fitness Landscape:

  • When DFEs are converted to unscaled selection coefficients ($s$), the distributions are broadly similar across all 38 primate subspecies.
  • There is no significant association between $N_e$ and unscaled selection coefficients, supporting the hypothesis of a conserved fitness landscape.

• Impact of Effective Population Size ($N_e$):

  • In scaled units ($S$), there is a strong positive correlation between $N_e$ and the fraction of strongly deleterious mutations ($S \in (-\infty, -10]$).
  • Conversely, there is a negative correlation between $N_e$ and the fraction of moderately/slightly deleterious and effectively neutral mutations.
  • Interpretation: This aligns with the Nearly Neutral Theory. Larger populations purge slightly deleterious mutations more efficiently, shifting the observed distribution of segregating variants toward the strongly deleterious tail.

• Dominance Effects:

  • Assuming mutations are partially recessive ($h < 0.5$) shifts the inferred DFE toward more strongly deleterious values (to compensate for masking in heterozygotes).
  • However, this shift is consistent across species. Therefore, comparative conclusions regarding DFE conservation remain robust regardless of the specific dominance assumption.
  • Simulations confirm that $h$ is poorly identifiable from SFS data alone due to strong non-identifiability with selection strength ($S$).

• Beneficial Mutations:

  • A positive correlation was found between $N_e$ and the proportion of fixed beneficial substitutions ($\alpha$), though this signal is sensitive to the DFE parameterization and SFS sample size.
  • The proportion of segregating beneficial mutations ($p_b$) is highly variable and difficult to estimate accurately from polymorphism data alone.

• Simulation Insights:

  • Demographic corrections using nuisance parameters successfully recover the true DFE shape even under strong bottlenecks and expansions.
  • Inference of the full DFE (including beneficial mutations) requires larger sample sizes ($n \ge 20$) and high SNP counts ($>10^4$) to distinguish weakly deleterious from weakly beneficial mutations.

5. Significance

This study provides a critical methodological and conceptual framework for comparative genomics. By explicitly distinguishing between scaled ($N_e s$) and unscaled ($s$) selection coefficients, the authors demonstrate that the apparent diversity in DFEs across primates is largely an artifact of differing effective population sizes rather than fundamental changes in the fitness landscape.

The findings reinforce the Nearly Neutral Theory as the primary driver of interspecific variation in polymorphism patterns. Furthermore, the development of tools to model dominance and the validation of inference under complex demography offer a more robust toolkit for future studies aiming to detect subtle, genuine changes in fitness landscapes across the tree of life. The work suggests that the primate fitness landscape is remarkably stable, with evolutionary differences arising primarily from how efficiently different populations can "see" and purge deleterious mutations based on their population size.