BGPT: Paper Review: Correlational selection in the age of genomics

Explore by Goal

Quick Explanation Copied

What this paper does

It synthesizes correlational selection theory and empirical/genomic evidence, focusing on how selection on trait combinations can reshape genetic correlations, pleiotropy, linkage disequilibrium, and genomic architecture (e.g., inversions/supergenes), while outlining practical genomic approaches to test these ideas.
Paper:

Long Explanation

Paper Review (visual-first): “Correlational selection in the age of genomics”

Nature Ecology & Evolution • Review Article • 15 Apr 2021 • Svensson et al.

One-sentence summary

The paper synthesizes multivariate/quadratic selection theory and genomic signatures to argue that correlational selection can reshape genetic covariance (G), mutational covariance (M), and genomic architecture—but notes that direct causal genomic evidence is still limited.

What counts as “correlational selection” (operational core)

Definition: selection where an individual’s expected fitness for one trait depends on that individual’s values for other traits—captured as interactions on the fitness surface.
Estimation: quadratic models of fitness include off-diagonal quadratic terms; non-zero off-diagonals indicate correlational selection.
Historical measurement framework: the Review explicitly connects this operationalization to classic multivariate selection measurement/fitness-surface estimation.

Visual map (conceptual, not data-fitting)

The paper’s central causal logic can be summarized as: fitness-surface interactions → correlational selection → changes in G and/or M via linkage disequilibrium & pleiotropy (± epistasis) → downstream genomic architecture (LD blocks, inversions/supergenes, recombination suppression) and organismal outcomes (integration, evolvability, plasticity, modularity).

Diagram is an interpretive synthesis of the causal chain described in the Review; it is not fitted to any quantitative data.

Where the Review is strongest

1) Clear mathematical operationalization + bridge to genomics

The paper does a good job keeping correlational selection anchored to an estimable quantity (quadratic fitness regression with γ off-diagonals) while mapping it to quantitative-genetic objects (G and M) and genomic architecture mechanisms.

2) Explicit acknowledgment of competing evolutionary forces

A key skepticism built into the narrative: genetic correlations can be shaped by mutation, drift, changing optima, and ecology—so correlational selection is not the sole driver. This is important because genomic signatures of selection are famously confounded by demographic history.

3) Practical “genomic signatures” menu (what would be evidence)

Box 2 lays out multiple routes—selective sweeps/polygenic allele shifts, linkage disequilibrium and age-class LD, recombination suppression clues (supergenes/rearrangements), and experimental validation via crosses/introgression/evolve-and-resequence.

Critical appraisal (skeptical, evidence-weighted)

A) Most genomic “consistency” is not direct causality

The Review is candid that direct evidence for long-term correlational selection shaping specific genomic architecture is scarce; much of the genomic narrative is inferential and compatible with multiple processes (drift, population structure, historical selection, and mutation/epistasis patterns).

B) LD vs pleiotropy vs epistasis: inference identifiability

The Review treats linkage disequilibrium and pleiotropy as two basic genomic causes of genetic covariances; however, from observational data, distinguishing them can be under-identified unless one has additional structural/genomic context (e.g., recombination suppression patterns, rearrangement breakpoints, or functional validation).

C) “Quadratic approximation” limitations for rugged fitness landscapes

The Review explains that quadratic models approximate fitness surfaces but may be inadequate if the true surface is rugged with multiple peaks/valleys; it notes non-parametric spline methods can capture complexity but may not yield coefficients integrated into evolutionary change equations.

D) Review bias: narrative synthesis, uneven empirical footing

As a Review article, its inferences are only as strong as the underlying empirical studies. That means taxonomic coverage and methodological diversity are likely uneven; and because correlational selection is hard to measure and sensitive to measurement choices, publication/reporting biases could matter (e.g., the field may over-represent “clear” multivariate interactions).

What would most strongly disprove or revise the Review’s program?

Demonstrate repeated cases where trait-combination selection exists/was inferred, yet there is no accompanying shift in genetic covariance structure consistent with alignment predictions (i.e., selection does not “write” into G in the expected directions).
Find genomic datasets where LD/recombination suppression and pleiotropic architecture appear but are better explained by demographic history or other selection regimes—i.e., correlational selection does not uniquely account for observed genomic structure.
Show that genotype–phenotype–fitness integration is not reproducible across systems, or that the implied G/M signatures do not track across life cycles.

Actionable reading guide (what to focus on next)

Topic	Why it matters	Critical skepticism question
γ-matrix / fitness-surface estimation	Determines whether correlational selection is actually present/quantified.	Is the quadratic approximation adequate for the surface shape?
G vs M evolution	Links selection to multivariate inheritance and mutational input.	Could drift/mutation dominate observed covariance changes?
LD blocks / inversions / supergenes	Proposed genomic mechanism for long-lived trait covariance.	Are genomic signatures uniquely consistent with correlational selection?
Plasticity + developmental alignment	Could recast “constraints” as context-dependent evolvability.	Are observed changes genetic (G) or environment-driven (plasticity)?

Table is an interpretive checklist derived from the Review’s organizing claims about fitness-surface estimation, G/M, genomic mechanisms, and open gaps.

Author reviews on BGPT

Open each to compare independent expert-style critique.

Feedback:

Updated: April 09, 2026