Most new mutations that change a protein do surprisingly little damage individually, reducing an organism’s survival or reproductive success by only a small fraction. Yet these mildly harmful changes vastly outnumber the rare mutations that are either lethal or beneficial, and their collective weight shapes how populations evolve, how diseases persist, and how species adapt to new challenges. The probability landscape describing all these possible fitness outcomes is called the distribution of fitness effects, or DFE, and understanding its shape turns out to be one of the most consequential problems in evolutionary biology.
What the Typical DFE Looks Like
Picture a curve that peaks sharply near zero harm and then trails off toward increasingly damaging effects. That is roughly what researchers find when they study new mutations in protein-coding genes. A landmark analysis of amino acid-changing mutations in humans showed that the data fit well to a particular statistical shape in which more than half of all new harmful mutations reduce fitness by somewhere between a thousandth and a tenth, placing them in the “mildly deleterious” category.1PubMed Central. The distribution of fitness effects of new deleterious amino acid mutations in humans In plain terms, most harmful mutations are not catastrophic. They are small drags on survival or reproduction, individually almost invisible to natural selection, yet collectively significant.
At the extremes, the picture is more dramatic. A fraction of mutations are lethal or so damaging that they are quickly weeded out. And an even smaller fraction are beneficial, giving the organism a slight edge. The overall shape is heavily lopsided: harmful mutations far outnumber helpful ones, and among the harmful ones, mild effects dominate over severe ones. This asymmetry has deep consequences for everything from the pace of adaptation to the accumulation of genetic disease.
Where Beneficial Mutations Fit In
Beneficial mutations are rare, and their fitness effects tend to follow a distinctive pattern. Theoretical work drawing on extreme value theory predicts that the distribution of fitness effects among beneficial mutations should be approximately exponential, meaning that most helpful mutations provide only a tiny boost, while large-effect beneficial mutations are exponentially rarer.2Genetics. The Distribution of Fitness Effects Among Beneficial Mutations This prediction has held up across multiple modeling frameworks, including phenotype-based models that treat organisms as occupying a multidimensional landscape of traits.3PubMed. The distribution of fitness effects among beneficial mutations in Fisher’s geometric model of adaptation
The exponential pattern matters practically because it tells us something about the raw material available for adaptation. When an organism faces a new stress, the mutations that help it survive will mostly be modest improvements, not dramatic leaps. Large-effect beneficial mutations do occur, but they are the exception. Adaptation, in most cases, proceeds through the accumulation of many small gains rather than a few transformative ones.
Empirical tests using bacteriophage and bacterial populations have generally supported this exponential shape, though the fit is not always perfect. Some studies have found evidence that the distribution’s tail is slightly heavier or lighter than a pure exponential would predict.4PubMed Central. Testing the extreme value domain of attraction for distributions of beneficial fitness effects The exponential remains the standard working assumption in the field, but researchers treat it as a useful approximation rather than a law of nature.
Viruses Break the Mold
Not every organism shows the same DFE shape. Experiments on five different virus species revealed a striking bimodal distribution, with one peak clustered near neutrality (mutations that barely change fitness) and another peak at lethality (mutations that destroy the virus entirely).5PubMed Central. A biophysical protein folding model accounts for most mutational fitness effects in viruses The middle ground of moderately harmful mutations was relatively empty.
This bimodal pattern makes sense when you consider how compact viral genomes are. Viruses carry very few genes, and each protein they encode is under intense functional constraint. A mutation either lands in a spot that can tolerate the change (producing a nearly neutral effect) or it disrupts the protein’s ability to fold and function (producing a lethal one). The researchers found that a simple biophysical model of protein folding could account for this pattern, suggesting that the shape of the DFE is partly a reflection of the physical chemistry of the proteins involved.
This finding highlights an underappreciated point: the DFE is not one universal curve. Its shape depends on the organism, the gene, and the type of mutation in question. What holds for a compact RNA virus does not necessarily hold for a large-genomed vertebrate with extensive regulatory DNA.
Coding Regions Versus Regulatory DNA
Even within a single organism, the DFE varies depending on which part of the genome you examine. A study of fruit fly populations compared the fitness effects of new mutations in protein-coding regions to those in functionally important non-coding (regulatory) regions. The regulatory mutations turned out to be predominantly moderately deleterious, whereas coding-region mutations skewed more heavily toward strongly deleterious effects.6PubMed Central. The distribution of fitness effects of new mutations in regulatory regions of the D. melanogaster genome
This distinction has practical consequences for how genetic variation accumulates. Mildly and moderately harmful mutations are harder for natural selection to eliminate efficiently, especially in small populations. Regulatory mutations, by falling into this intermediate zone, may linger in a population for many generations. Strongly deleterious coding mutations, by contrast, tend to be purged quickly. The upshot is that the regulatory genome may be a larger reservoir of “hidden” genetic damage than the protein-coding genome, even though protein-coding mutations tend to get more attention in disease genetics.
How Researchers Estimate the DFE
You cannot directly observe the fitness effect of every possible mutation in a living organism, so researchers use several indirect strategies. Each has strengths and blind spots, and the estimates they produce do not always agree perfectly.
One classic approach is the mutation accumulation experiment. Populations are maintained at very small sizes, often passing through a single individual each generation, so that natural selection has almost no opportunity to filter out harmful mutations. After many generations, the accumulated mutations are analyzed for their effects on traits like survival, fertility, or growth rate. Because selection is essentially turned off during the experiment, the mutations that pile up should represent the raw spectrum of what spontaneous mutation produces.7Nature. Selection in a growing colony biases results of mutation accumulation experiments These experiments were pioneered decades ago and provided some of the earliest estimates of mutation rates and average fitness effects, though they are labor-intensive and can take years to complete.8Oxford Academic. Old Trade, New Tricks: Insights into the Spontaneous Mutation Process from the Partnering of Classical Mutation Accumulation Experiments with High-Throughput Genomic Approaches
A more modern strategy uses patterns of genetic variation in natural populations. By comparing the frequency spectrum of mutations at sites under selection to those at neutral sites, researchers can infer the shape of the DFE without ever directly measuring fitness. The challenge is that past population-size changes, such as bottlenecks or expansions, also shift allele frequencies in ways that can mimic or mask the signature of selection. Failing to account for demographic history can bias DFE estimates substantially.9Genetics. Joint Inference of the Distribution of Fitness Effects of Deleterious Mutations and Population Demography Based on Nucleotide Polymorphism Frequencies Modern methods try to jointly estimate both the DFE and the population’s demographic history from the same data, a statistically demanding task that continues to be refined.
Deep mutational scanning, a high-throughput laboratory technique, offers yet another window. Researchers systematically create thousands of single-mutation variants of a protein and measure how each one performs in a controlled assay. Recent work has begun combining data from multiple proteins to build more general models of how mutations affect protein fitness landscapes.10PubMed. Learning protein fitness landscapes with deep mutational scanning data from multiple sources These experiments provide an extraordinarily detailed view of a single protein’s DFE but may not capture the full range of fitness effects that matter in a whole organism living in its natural environment.
How Environment Reshapes the DFE
A mutation’s fitness effect is not a fixed property stamped into the DNA. The same mutation can be mildly harmful in one environment and neutral or even beneficial in another. Researchers tested this by measuring the fitness effects of dozens of individual mutations in fruit flies across different environmental conditions and genetic backgrounds. They found that both the environment and the genetic background changed how individual mutations performed. When they looked at the overall shape of the DFE, however, the genetic background mattered less than the environment: the average severity and the spread of fitness effects both shifted when the environment changed.11PubMed. Sensitivity of the distribution of mutational fitness effects to environment, genetic background, and adaptedness: a case study with Drosophila
Work on plant viruses has produced even more dramatic examples. When tobacco etch virus was grown in its natural host family, the DFE was dominated by deleterious mutations. But when the same virus was transferred to distantly related hosts, average fitness dropped sharply and the fraction of beneficial mutations grew significantly.12PLoS Genetics. Effect of Host Species on the Distribution of Mutational Fitness Effects for an RNA Virus In the unfamiliar hosts, mutations that would have been neutral or harmful in the original environment suddenly had a chance of improving fitness, simply because the virus was so poorly adapted to begin with.
This context-dependence is not just an academic curiosity. It means that predictions about how a pathogen will evolve in a new host, or how a crop pest will respond to a changed climate, depend critically on knowing the DFE in the relevant environment, not just in the laboratory conditions where it was first characterized.
How Mutations Interact With Each Other
The DFE is usually described as if each mutation acts independently, but real genomes contain many mutations simultaneously, and their combined effects do not always add up neatly. Sometimes two mildly harmful mutations together are worse than the sum of their individual harms. Sometimes the combination is less bad than expected. These non-additive interactions are collectively called epistasis, and they can substantially reshape the effective DFE that a population experiences over time.
The amount of epistasis depends on both the fitness landscape itself and the types of mutations being considered. New mutations, standing variation already circulating in a population, and mutations that have become fixed all sample different parts of the landscape, and the degree of interaction among mutations varies accordingly.13PubMed Central. The distribution of epistasis on simple fitness landscapes One useful approach for quantifying these interactions looks at how the correlation in fitness effects between pairs of mutations changes as the genetic distance between them increases. The way this correlation decays can reveal the underlying structure of the fitness landscape, essentially telling researchers whether the terrain is smooth, rugged, or somewhere in between.14PubMed. Measuring epistasis in fitness landscapes: The correlation of fitness effects of mutations
Fisher’s geometric model, one of the most influential theoretical frameworks in the field, naturally generates epistatic interactions as a consequence of its multidimensional structure.15PubMed Central. The Utility of Fisher’s Geometric Model in Evolutionary Genetics In this model, an organism’s fitness depends on how close it is to an optimal combination of many traits. A mutation that moves the organism closer to the optimum in one dimension may push it further away in another, and whether that tradeoff is net positive or net negative depends on where the organism currently sits. The model predicts that beneficial mutations should become rarer and smaller in effect as a population approaches its optimum, a prediction that lines up with empirical observations in several microbial systems.
Dominance and the Hidden Burden of Mutations
In organisms like humans, most genes exist in two copies. A new mutation initially appears in just one copy, and its fitness effect depends partly on how much damage the single mutant copy causes when the other copy is still normal. Geneticists describe this with a dominance coefficient: fully recessive mutations cause harm only when both copies are mutant, while fully dominant mutations cause harm even with one normal copy present.
A consistent finding across species is that strongly deleterious mutations tend to be more recessive than mildly harmful ones. Data from yeast gene knockouts showed this clearly: alleles with large effects when both copies were mutant tended to have smaller effects when only one copy was affected.16PubMed Central. Inferences about the distribution of dominance drawn from yeast gene knockout data Studies fitting models to human genomic data have reached a similar conclusion, finding that the data are consistent with strongly deleterious mutations having very low dominance coefficients, around 0.05.17PubMed Central. Constraining models of dominance for nonsynonymous mutations in the human genome A broader review of dominance estimates confirmed this inverse relationship between the severity of a mutation and how dominant it is, with several studies showing that the dominance coefficient shrinks as the selection coefficient grows.18Genome Biology and Evolution. Revisiting Dominance in Population Genetics
This pattern has a profound consequence for genetic load, the total burden of harmful genetic variation a population carries. Because strongly damaging mutations tend to be hidden in their single-copy state, they can persist in a population far longer than they would if their effects were immediately visible to selection. Carriers of one copy suffer little or no penalty, so the mutation quietly spreads before two carriers happen to have offspring who inherit both copies. This is exactly the pattern seen in many inherited diseases in humans, where carrier frequencies for severe conditions can be surprisingly high.
Population Size and the Efficiency of Selection
Whether a mildly harmful mutation is eliminated by selection or drifts to high frequency depends heavily on population size. In large populations, even small fitness differences are “visible” to natural selection, and mildly deleterious mutations tend to be weeded out. In small populations, random chance dominates, and many mildly harmful mutations behave as though they were neutral, accumulating as if selection were not there at all.
A comparative study across a wide range of species found that organisms with smaller long-term effective population sizes showed reduced efficiency of natural selection, measured as a higher ratio of harmful to neutral genetic variation.19PubMed Central. A Comparative Analysis of Long-Term Effective Population Sizes Across Eukaryotes This finding has direct implications for conservation biology. Small, isolated populations, whether of endangered species or fragmented plant populations, are expected to accumulate mildly deleterious mutations at a faster rate than their larger counterparts. Over time, this accumulation can reduce the population’s overall fitness, a process sometimes described as genetic load contributing to an extinction vortex.
Every organism carries some burden of disadvantageous genetic variants.20PubMed Central. Genetic load In large, well-connected populations, selection keeps this burden in check. In small or declining populations, the burden grows, potentially lowering survival, reproductive success, and resilience to environmental change. Understanding the DFE is therefore not just an abstract exercise; it directly informs predictions about which populations are most vulnerable and how quickly genetic deterioration can proceed.
Antibiotic Resistance and Compensatory Evolution
One of the most tangible applications of DFE research involves antibiotic resistance. When a bacterium acquires a resistance mutation, that mutation often comes with a fitness cost: the bacterium grows more slowly or competes less well in the absence of the drug. But bacteria do not simply sit with that cost. They evolve compensatory mutations that restore fitness while keeping the resistance.
Experiments with antibiotic-resistant bacteria have shown that these compensatory beneficial mutations arise at high rates and produce average fitness gains of roughly two to four percent, depending on the cost of the original resistance mutation. When the initial cost was higher, the compensatory mutations tended to have larger effects, consistent with the theoretical prediction that organisms further from their fitness optimum have access to bigger beneficial steps.21PubMed Central. Cost of Antibiotic Resistance and the Geometry of Adaptation This is a concrete example of how the DFE changes with context: the landscape of available beneficial mutations shifts depending on how far from optimal the population currently is.
The practical implication is sobering. If compensatory mutations are abundant and arise quickly, the fitness cost of resistance may be a poor barrier to the spread of resistant bacteria. The DFE of compensatory mutations essentially determines how long a resistance mutation stays “expensive” for the bacterium, and the answer, in many cases, is not very long.
Directed Evolution and Biotechnology
Outside of natural populations, the DFE plays a role in protein engineering and directed evolution, where researchers mutate proteins in the laboratory and select for improved function. The strategy a researcher should use to find the best variants depends on the shape of the DFE for that protein in that context.
Modeling work has shown that when fitness effects are highly variable from one mutant to the next, a diversification strategy (testing many different variants rather than repeatedly improving the current best one) tends to produce better outcomes. Larger population sizes and more rounds of selection also favor diversification, because they give researchers a better chance of sampling the rare high-fitness variants hidden in the right tail of the distribution.22PubMed Central. Effects of selection stringency on the outcomes of directed evolution In other words, the shape of the DFE dictates the optimal experimental design. If most beneficial mutations are small and the distribution is tightly clustered, a greedy hill-climbing approach works fine. If the distribution is broad with a heavy tail, casting a wider net pays off.
This connection between evolutionary theory and laboratory practice is a growing area. As deep mutational scanning data accumulate for more and more proteins, researchers are building computational models trained on these real DFEs to predict which mutations in a new protein are likely to be beneficial, sometimes without ever testing them directly in the lab. The better the community’s understanding of the DFE’s typical shape and how it varies, the more efficiently these engineering efforts can be guided.
Why Getting the Shape Right Matters Beyond Biology
Estimates of the DFE feed directly into models used for population genetic inference, conservation management, and even forensic genetics. When population geneticists try to reconstruct the demographic history of a species from its DNA, they need to separate the signature of selection from the signature of population-size change. Getting the DFE wrong distorts those demographic estimates, potentially leading to incorrect conclusions about when bottlenecks occurred or how populations expanded.9Genetics. Joint Inference of the Distribution of Fitness Effects of Deleterious Mutations and Population Demography Based on Nucleotide Polymorphism Frequencies
In conservation genetics, the DFE shapes recommendations about how to manage endangered populations. If a species’ DFE is dominated by mildly deleterious mutations, maintaining larger populations and encouraging gene flow between fragments becomes especially important, because those are the mutations most sensitive to population size. If, instead, the DFE is dominated by strongly deleterious or lethal mutations, selection will purge most of the damage regardless of population size, and the conservation calculus shifts. The answer is rarely one or the other in pure form, but leaning heavily in one direction changes what interventions are most cost-effective.
In human medical genetics, DFE estimates inform predictions about the proportion of newly arising mutations that contribute to complex diseases. Because the DFE for human coding mutations skews toward mild effects, much of the genetic variation underlying common diseases may consist of individually weak-effect variants, each reducing fitness only slightly. This is consistent with the observation that genome-wide association studies for most complex traits turn up hundreds or thousands of associated genetic variants, each contributing a tiny fraction of the overall risk. The shape of the DFE, in a sense, predicted the architecture of complex disease long before the data confirmed it.