BGPT: Paper Review: Quantitative genetics of microbiome-mediated traits

Explore by Goal

Quick Explanation Copied

Paper focus (Week et al., 2025)

Builds a quantitative-genetics framework for microbiome-mediated host traits by: (i) partitioning additive host-microbiome variance into host, microbe, and gene–microbe covariance components ; (ii) classifying microbes by ancestral concordance (lineal / non-lineal / novel) ; (iii) deriving a generalized breeders equation and microbiome-informed transmissibility indices .

Long Explanation

Paper review: Quantitative genetics of microbiome-mediated traits

Week et al. (Evolution; DOI: 10.1093/evolut/qpaf171) propose a quantitative-genetics extension that (1) partitions host trait variance into host-genetic and microbiome-mediated components, (2) partitions microbiome into ancestral-concordance components (lineal/non-lineal/novel), and (3) derives microbiome-aware transmissibility and a generalized breeders equation for predicting host selection responses.

What’s new here?

Variance partitioning includes microbial additive effects and gene–microbe covariance in an explicitly quantitative-genetic way.
Microbes are classified by lineal vs non-lineal vs novel ancestry concordance to decide which microbial components are plausibly “selective factors.”
A generalized breeders equation is introduced for microbiome-mediated traits: z = (G_A + M_ψA + C_ψA) β.

1) Framework map (how the pieces connect)

The paper’s logic is: start from Fisher-style additive decomposition of continuous trait variation and the modern heritability interpretations ; then generalize to host genotype g and host microbiome m via genotypic-microbic values z_gm .

Scientific note (skeptical): This diagram is a conceptual “wiring map,” not a claim about biological universality. The paper itself emphasizes that microbial transmission/fidelity and within-host dynamics are complex and are approximated/simplified in the simulation components.

2) Key equations (variance decomposition + selection prediction)

The paper’s central additive decomposition for host genotype g and microbiome m is P_A = G_A + M_A + C_A, where:

G_A = additive genetic variance across host allele-count covariances
M_A = additive microbial variance across microbe relative-abundance covariances
C_A = additive gene–microbe covariance term (covariation between allele counts and microbe abundances)

Component	Symbol(s)	What it means (from the paper)	Why it matters for inference
Additive variance	P_A	Additive component of trait variance explained jointly by host genotype and host microbiome factors.	Provides the measurable “decomposition target” analogous to additive genetic variance in classical quantitative genetics.
Host additive variance	G_A	Additive genetic variance term, built from allele-count variances and covariances.	Establishes baseline “classical” expectation; can be benchmarked against standard heritability approaches (with the usual caveats about interpretation).
Microbial additive variance	M_A	Additive microbial variance term, built from microbe relative-abundance variances/covariances.	Is necessary but not sufficient for selection prediction: the paper stresses transmissibility/ancestral concordance matter for evolutionary response.
Gene–microbe covariance	C_A	Additive gene–microbe covariance across hosts (covariation between allele counts and microbe abundances).	Can be positive or negative and therefore can bias “total” transmissibility estimates if ignored.
Generalized breeders equation	z = (G_A + M_ψA + C_ψA) β	Response to host selection depends on host additive genetic variance plus only the microbial additive variance/covariance from selective (transmissible + selection-aligned) microbial components.	Targets the frequent pitfall: variance explained by microbiome data can systematically overestimate selection response if non-transmissible components are included.

3) The ancestral-concordance classification: strengths + blind spots

What the paper claims: Microbes can be grouped into: (1) Lineal microbes whose ancestry remains associated with host lineages (co-diverging / co-transmitted over multiple generations), (2) Non-lineal microbes whose ancestry is partially concordant with host ancestry, and (3) Novel microbes whose ancestry is outside the host population and thus (in their model) not inherited via host lineages.

Why this helps (inference)

It offers a mechanism-agnostic criterion: the classification is based on ancestry patterns rather than prescribing specific transmission rules up front.
It directly addresses the key evolutionary question: which microbial components can be selective factors contributing to intergenerational response.

Potential blind spots

Discrete taxon categories (lineal/non-lineal/novel) may not match within-taxon heterogeneity: the paper itself notes that taxa are mixtures of microbes with different ancestral concordance degrees, and future work would need distributions rather than hard partitions.
Empirical identifiability of ancestral concordance is non-trivial: it depends on having lineage-traced or genomic ancestry signals for microbes and hosts, and on mapping “microbial individuals” to sequencing features. The paper motivates this as potentially measurable but does not provide a concrete universal inference pipeline.
Dependence on model timing between selection events and microbial acquisition: the contribution of non-lineal microbes depends on whether they are sourced from pre- vs post-selected parents in their simulation.

4) What the simulations show (and what they don’t)

Design (high-level): The paper simulates host genetics and microbiomes with explicit inheritance modes: lineal microbes are transmitted around mid-parent values (Poisson around mid-parents), novel microbes are acquired independently each generation (iid Poisson), and non-lineal microbes are acquired from either pre-selected or post-selected parental populations.

Qualitative figure summary (from the paper)

Fig. 3: selection-response dynamics differ depending on whether only genes (G), genes + lineal microbes (GL), genes + lineal + non-lineal microbes (GLN), or genes + lineal + non-lineal + novel microbes (GLNV) are included; the paper reports that novel microbes mostly become hidden/irrelevant in correct predictions but can distort variance-based predictions when included.
Fig. 4: correlations between allele counts and microbe abundances evolve differently for lineal vs non-lineal/novel microbes depending on whether non-lineal sources are pre- or post-selected.
Fig. 5: predicted vs observed changes in host mean trait show overestimation when including novel microbes and/or certain non-lineal components; best agreement occurs when including host genes + lineal microbes and excluding novel microbes, and when including certain non-lineal components consistent with post-selected sourcing.

Skeptical interpretation

The results are simulation-based; they test internal consistency of the framework’s logic under controlled assumptions, not general biological realism. The paper explicitly flags this with limitations (e.g., ignoring within-host microbiome dynamics and ecological assembly).
The “novel microbes inflate variance but not response” pattern is logically plausible under the paper’s selective-factor restriction. However, real microbiomes may include taxa with long persistence, multiple generations of overlap, and measurement/feature mapping issues—any of which could change the effective selective component set.

5) Relevance for association studies (MWAS/GWAS): what to be careful about

The paper argues that naïve MWAS/GWAS-style interpretations can fail because associations reflect ecological dynamics and not necessarily transmissible, selection-aligned components.

Additional skeptical note (data modality)

Because microbiome data are compositional in many standard pipelines, relative-abundance heritability can be biased by inter-taxa dependencies. This doesn’t directly invalidate the Week et al. theoretical framework, but it raises the practical question of how one would estimate ω_m, M_A, and C_A from sequencing features without compositional artifacts. .

6) How to most directly falsify (empirically) the paper’s predictions

The cleanest empirical test would try to create a setting where:

Host lineages are trackable (pedigree/genomic relatedness is known).
Microbial ancestry concordance can be estimated at least approximately (e.g., via microbial genomic linkage/marker persistence or lineage-tracing approaches).
Selection on a host trait is imposed and repeated across generations.
You measure whether observed trait response matches the generalized breeders-equation prediction only when microbes corresponding to “selective factors” (lineal and appropriately sourced non-lineal) are included, and not when “novel” components are included.

This falsification logic follows directly from the simulation comparisons reported in the paper (overestimation when novel microbes are included; prediction matches when transmissible/selection-aligned components are used).

Practical takeaway

If you want microbiome signals to inform evolutionary prediction (not just correlation), you must treat microbiome features as selective factors with transmissibility constrained by ancestral concordance, and you must include gene–microbe covariance terms rather than collapsing everything into “microbial variance explained.”

Author reviews (start with any)

Feedback:

Updated: April 18, 2026