TL;DR
This paper argues that standard population genetics models, which assume simple "Kingman coalescent" scenarios, are insufficient for accurately studying viral evolution. Instead, it highlights the critical need to incorporate "multiple merger coalescent" (MMC) events and "background selection" (BGS) to account for the unique characteristics of viral populations, such as skewed offspring distributions and high mutation rates.
What Problem Does This Paper Address?
Population geneticists often use models based on the Kingman coalescent to infer evolutionary history. This model assumes that in large populations, only two lineages can merge (coalesce) at any given time, and that offspring distributions are Poisson-shaped. However, viruses exhibit several traits that violate these assumptions:
- Skewed Offspring Distributions: Viruses often have highly variable reproductive success, meaning a few individuals can produce a disproportionately large number of offspring. This is unlike the more even distribution assumed by the Kingman model.
- Frequent Bottlenecks and Strong Selection: Viral transmission and host immune responses often involve severe population bottlenecks and strong selective pressures, further distorting genetic variation.
- High Mutation Rates and Compact Genomes: Viruses, especially RNA viruses, have very high mutation rates and relatively small, compact genomes. This makes them susceptible to the accumulation of deleterious mutations and the effects of background selection.
Failing to account for these viral-specific factors can lead to significant misinterpretations of evolutionary parameters like population size, mutation rates, and selection strengths.
Methods
The paper doesn't introduce new computational methods but rather advocates for the use of more sophisticated theoretical frameworks:
- Multiple Merger Coalescent (MMC) Models: These models generalize the Kingman coalescent by allowing more than two lineages to merge simultaneously. This better reflects scenarios with skewed offspring distributions, such as those seen in viruses. Various MMC models are discussed, including the $\psi$-coalescent, which has a direct biological interpretation related to reproductive events.
- Background Selection (BGS) Modeling: BGS describes how the purging of deleterious mutations in a region of low recombination can reduce genetic diversity and affect linked neutral variation. The paper emphasizes that accurately modeling BGS, especially for weakly deleterious mutations, is crucial for viruses with high mutation rates and compact genomes.
- Simulation-based Inference: The authors suggest that for complex scenarios involving non-equilibrium demography, selection, and BGS, direct simulation (e.g., using Approximate Bayesian Computation - ABC) is often more appropriate than analytical methods or simple parameter rescaling.
Datasets
Not discussed in the paper. The paper focuses on theoretical frameworks and their application to understanding viral population genetics, rather than analyzing specific empirical datasets.
Key Insights and Ideas
- Kingman Coalescent is Often Insufficient for Viruses: The paper strongly argues that the standard Kingman coalescent model, while useful for many organisms, is a poor fit for viruses due to their unique life history and evolutionary dynamics.
- Skewed Offspring Distributions Drive MMC: The highly variable reproductive success of viruses naturally leads to scenarios where multiple lineages can coalesce at once, making MMC models a more appropriate choice.
- Background Selection is Pervasive in Viruses: High mutation rates and compact genomes mean that deleterious mutations can accumulate and be purged, leading to significant background selection effects that impact neutral variation.
- Mis-inference is a Real Risk: Ignoring MMC and BGS can lead to drastically incorrect estimates of key population genetic parameters, such as effective population size ($N_e$) and mutation rates.
- Model Choice is Crucial and Context-Dependent: The "best-fit" coalescent model for a virus depends on its specific biology and the scale of the evolutionary process being studied (e.g., intra-host vs. inter-host).
Key Results
The paper presents theoretical arguments and synthesizes existing research to demonstrate:
- Skewed offspring distributions: Can lead to elevated linkage disequilibrium (LD) and skewed estimates of population differentiation ($F_{ST}$), even with high recombination rates.
- MMC models: Produce distinct patterns in genealogies (more star-like) and site frequency spectra (SFS) compared to Kingman coalescent models, often showing an excess of low-frequency and high-frequency variants.
- Background selection: Can lead to underestimated effective population sizes if not properly accounted for, particularly when weakly deleterious mutations are prevalent.
- Example of Pacific Oysters: Eldon and Wakeley (2006) showed that ignoring skewed offspring distributions in oysters led to mutation rate estimates that were two orders of magnitude larger than those obtained using a more appropriate $\psi$-coalescent model.
Limitations and Future Work
The authors identify several key areas for future research:
- Developing Summary Statistics: There's a need for summary statistics that can effectively capture demographic features and processes in highly skewed offspring distributions, enabling efficient large-scale inference.
- Understanding Existing Statistics: Further investigation is needed into how commonly used statistics (like $F_{ST}$, SFS, and LD) behave under MMC models.
- Tailoring MMC Models to Virus Scales: Determining which specific MMC models are best suited for different scales of viral evolution (intra-host, inter-host, global) is crucial.
- Accounting for BGS: Future work should focus on directly simulating weakly deleterious mutations to account for BGS effects on the SFS, rather than relying on simple rescaling of $N_e$.
- Improving Recombination Maps: Accurate recombination maps for viral genomes are essential for inferring BGS.
- Integrated Models: Developing models that jointly account for non-equilibrium demography, positive selection, and BGS would allow for a more comprehensive estimation of all relevant parameters.
- Extending Methods: Applying methods developed for other taxa (like establishing null expectations for BGS) to virus populations is a promising avenue.
- Simulation-based Inference: The development of efficient simulation frameworks for MMC models will be key for implementing ABC inference and gaining a better understanding of these processes.
Summary updated