Environmental Disorder Regulation of Invasion and Genetic Loss

Mahdi Sasar; V\'ictor M. P\'erez Garc\'ia; Youness Azimzade

REVIEW 3 major objections 4 minor 28 references

Environmental Disorder Regulation of Invasion and Genetic Loss

T0 review · 3 major / 4 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Spatial fluctuations in movement speed regulate genetic loss during range expansion, with weak disorder slowing it and strong disorder accelerating it.

desk verdict Front-wandering scaling may be real, but the genetic-loss headline rests on a false homogeneous-extinction baseline. read the letter →

arxiv 1908.02532 v1 pith:IKZX7TEG submitted 2019-08-07 q-bio.PE physics.bio-ph

Youness Azimzade , Mahdi Sasar , Víctor M. Pérez García This is my paper · ORCID

classification q-bio.PEphysics.bio-ph MSC 92D1592D2535K5760H15

keywords rangeexpansiongeneticdriftreaction-diffusionequationspatialdisorderinvasionfrontheterozygositystochasticdiffusivityFKPP

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper asks whether randomness in the environment—specifically, spatial fluctuations in how fast individuals move—can change the fate of genetic diversity during a biological invasion. The authors model an invading population with the Fisher-Kolmogorov-Petrovsky-Piskunov (FKPP) equation, replacing the constant diffusion coefficient by $D_0(1+\xi f(x))$ with white-noise $f(x)$. They derive a Langevin equation for the front position whose effective diffusion coefficient is $D_C = \frac{1}{8}\,\xi^2\sqrt{RD_0}$, and they show numerically and analytically that the same noise controls genetic loss: weak fluctuations (small $\xi$) delay the extinction of a minority population and raise heterozygosity at the front, while strong fluctuations accelerate extinction. If true, environmental disorder is a regulator of genetic drift in its own right, independent of fitness differences or population-size fluctuations.

What carries the argument

The argument is carried by a perturbative treatment of the stochastic reaction-diffusion equation in a comoving frame. Writing the density as $C(\zeta,t)\approx C_0(\zeta+\eta(t),t)+\delta C_1(\zeta,t)$, with $C_0$ the solution of the unperturbed linearized equation, the authors project the noise term onto the zero eigenfunction $e^{\sqrt{R/D_0}\,\zeta}$ of the adjoint operator. This yields an equation of motion for the front displacement $\eta(t)$ and, after ensemble averaging $\langle f(x)f(y)\rangle=\delta(x-y)$, the effective diffusion coefficient $D_C$. For genetic loss, the key object is the difference field $\Upsilon=C_2-C_1$ and its squared integral $\kappa(t)=\int \Upsilon^2\,dx$; a Green-function calculation gives $\langle\Delta\kappa^2\rangle$, and direct simulation of the two-population equation supplies the heterozygosity $H=\langle\int C_1C_2\,dx\rangle$ that exhibits the non-monotonic dependence on $\xi$.

What would settle it

A direct test is to integrate the deterministic two-population PDE (13) with $\xi=0$, equal fitness and no demographic noise, starting from $C_1=0.49$, $C_2=0.51$: if the minority population does not go extinct asymptotically, the paper's reference behavior for 'genetic loss' is invalid. A complementary experiment would track heterozygosity in a microbial or cell range expansion on substrates with controlled diffusivity variance, looking for the predicted non-monotonic response with a minimum of genetic loss near $\xi\approx0.2$.

Watch

Extended reading notes

Core claim

The central discovery is that spatially random diffusivity acts non-monotonically on genetic loss during range expansion. In the paper's two-population model, starting from nearly equal densities $C_1=0.5-\epsilon$ and $C_2=0.5+\epsilon$, the minority population eventually disappears in a uniform environment; introducing disorder of amplitude $\xi$ delays that loss for $\xi\lesssim0.2$ and speeds it up for $\xi\gtrsim0.2$. The front itself performs an unbiased random walk, so the invasion border wanders as $\langle (X_F-\bar X_F)^2\rangle \sim 2D_C t$ with $D_C=\frac18\xi^2\sqrt{RD_0}$. The authors interpret this as evidence that regional differences in the intensity of genetic drift—such as those observed in human expansions and tumor mutational landscapes—can be produced by environmental heterogeneity alone.

Load-bearing premise

The load-bearing premise is that, in a perfectly uniform environment, the initially less abundant population is doomed to extinction; if that deterministic baseline is not accurate, the paper's claim that weak disorder delays genetic loss is measured against the wrong reference.

Editorial extensions

If this is right

An invasion front in a habitat with random motility performs an unbiased random walk, so the position variance grows linearly in time with coefficient $\frac18\xi^2\sqrt{RD_0}$; this can be tested by tracking range edges in experiments or field data.
Weak environmental disorder (roughly $\xi<0.2$) preserves genetic diversity at the advancing front by delaying the extinction of less abundant populations.
Strong disorder ($\xi>0.2$) does the opposite, accelerating genetic loss and making the loss spatially patchy even when fitness is equal.
Observed regional differences in human genetic loss intensity could be explained by differences in environmental variability rather than by distinct demographic histories alone.
In tumors, clonal diversity (and hence mutational heterogeneity) may be shaped by the physical disorder of the surrounding tissue through the same mechanism.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

Inference: a similar non-monotonic response should appear in two- and three-dimensional expansions, where the disorder variance replaces the one-dimensional $\xi^2$; the transition point may shift with dimension and with the correlation length of the noise.
Inference: the model predicts an experimentally accessible control variable—substrate heterogeneity in microbial range expansions—so the claim can be tested by growing populations on surfaces with engineered patches of different motility and measuring heterozygosity.
Inference: if this mechanism operates in humans, genetic-loss rates should correlate with measurable environmental variance (terrain ruggedness, resource patchiness) along the expansion path, not just with distance from the origin.
Inference: because the theory only requires a fluctuating diffusion constant, any process that modulates dispersal—seasonality, climate variability, or tissue stiffness—could play the same regulatory role, broadening the relevance beyond the paper's examples.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The paper has one genuinely useful piece—front wandering in a FKPP front with spatially quenched diffusivity noise—and one headline result that appears to rest on a numerical artifact. The claim that small disorder suppresses genetic loss and large disorder intensifies it is not supported by the model as written.

What is new and good: the effective front diffusivity D_C ~ xi^2 sqrt(R D0) is a parameter-free scaling law, numerically confirmed over a wide range. The zero-mode projection method is standard, but applying it to quenched diffusivity noise is a fair new application, and the simulation support for the exponents is solid.

The soft spots: first, the noise prescription is inconsistent. The model defines f as bounded uniform noise on [-1,1], while the analytics use Gaussian delta-correlated white noise with = delta(x-y). Those are not interchangeable, and the difference affects the constant in D_C. The step from Eq. (10) to (11) is also not shown, so the stated 1/8 prefactor is not derived in any transparent way.

Second, and more serious: the genetic-loss section claims that in a homogeneous environment C1 goes extinct because of its smaller initial value. But Eq. (14b) for Upsilon = C2 - C1 is linear, with no stochastic source and a nonnegative growth term. If Upsilon starts at 2epsilon, it remains strictly positive; there is no deterministic extinction mechanism. The zeta=0 curve in Fig. 3 likely shows extinction only because the numerics round flows to integer multiples of 10^-10, an artificial cutoff. That makes the baseline for the 'weak disorder delays extinction' story false, and the nonmonotonic heterozygosity curve is measured against the wrong reference. The model also has no demographic stochasticity, so 'genetic drift' in the usual sense is absent; environmental disorder acting identically on both populations cannot break the symmetry between equal-fitness types.

The front-wandering result may survive additional work, but the paper's central biological claim does not. A proper stochastic branching process or an analytically justified extinction threshold could revive the genetic-loss story, but as submitted it is one good scaling law attached to a broken headline.

Recommendation: send it to peer review if you want the authors to fix the baseline and re-examine the noise prescription. It deserves a serious referee, but I would expect heavy revision.

Referee Report

3 major / 4 minor

Summary. The manuscript studies a one-dimensional FKPP-type reaction-diffusion equation with a spatially fluctuating diffusion coefficient, D(x) = D0(1 + ξ f(x)), where f(x) is introduced as uniform noise in [-1,1]. The authors derive a Langevin description for the position of the invasion front and obtain an effective diffusion coefficient D_C = (1/8) ξ^2 sqrt(R D0) for front wanderings, which they support with numerical fits reproducing the scaling in ξ, R, and D0. The second part addresses genetic loss by considering two neutral populations with slightly different initial densities and analyzes how spatial disorder in diffusivity affects their coexistence. The paper claims that weak environmental fluctuations (small ξ) delay genetic loss, while strong fluctuations (large ξ) intensify it, and it interprets this as a potential explanation for regional differences in human genetic diversity and tumor heterogeneity.

Significance. If correct, the front-wandering result would be a useful quantitative contribution to the theory of reaction-diffusion fronts in heterogeneous media, and the genetic-loss result would be a striking and biologically relevant prediction. The numerical validation of the D_C scaling is, on its face, well executed: the paper reports scaling exponents and R² values, and the perturbation derivation follows a standard route. The genetic-loss claim, however, is the central novelty of the paper, and it rests on assumptions that contradict the model equations. Because the claimed regulatory role of environmental disorder on genetic loss is not established by the reported analysis, the overall significance of the paper is substantially diminished. The front-wandering part may be salvageable, but the genetic-loss claim as presented is not.

major comments (3)

[Model and Invasion Front Wanderings, Eq. (2) and Eq. (9)] The noise prescription is internally inconsistent: the model defines f as 'uniform white noise in the range [-1,1]', but the ensemble average leading to Eq. (9) uses <f(x)f(y)> = δ(x-y), which is the correlation of Gaussian white noise with unit variance, not of bounded uniform noise. For uniform noise in [-1,1] one would have <f(x)f(y)> = (1/3)δ(x-y), changing the prefactor in Eq. (11). The numerical fits in Fig. 2 compare only scaling exponents and do not verify the prefactor 1/8. The derivation of D_C is therefore not a parameter-free quantitative prediction for the simulated model.
[Genetic loss, paragraph before Fig. 3] The baseline for the genetic-loss claim is incorrect. The text states that for non-zero ε in a homogeneous environment C1 will be extinct 'due to its smaller initial value', but this contradicts Eq. (14b). The difference Υ = C2 - C1 satisfies a linear equation with diffusion and growth term R(1-C)Υ, with R(1-C) ≥ 0, so starting from Υ(x,0) = 2ε > 0 yields Υ(x,t) > 0 for all time. The minority population cannot go extinct in the deterministic continuum model. The ξ=0 extinction curve in Fig. 3(a) is therefore a numerical artifact, most plausibly the discretization rule that rounds all flows to integer multiples of 10^-10. The comparison showing that weak disorder 'delays' extinction has no valid baseline.
[Genetic loss, Eqs. (16)-(19)] The analytic derivation of noise-induced fluctuations in κ rests on an uncontrolled approximation. In Eq. (16) the authors neglect the term 2R∫ C Υ² dx by asserting that C≈1 only where Υ≈0 and C≈0 only where Υ≈1. In a traveling front, however, there is a finite interfacial region where C and Υ are both of order unity, and there is no separation of scales that makes the product negligibly small. Since this approximation is the basis for Eqs. (17)-(19) and for the claimed fluctuation <Δκ²>, the analytic support for the genetic-loss result is not established.

minor comments (4)

[Model section and Eq. (2)] The text writes f(x) as spatial noise but Eq. (2) uses f(x,t); the time dependence is not defined and is later dropped. Please clarify the notation.
[Model section] The phrase 'uniform white noise in the range [-1,1]' is ambiguous: it could mean uncorrelated uniform spatial noise, but 'white noise' usually implies Gaussian delta-correlated noise. The normalization and correlation function of f should be stated precisely.
[Fig. 3 caption] The caption uses 'adimensional time'; the correct term is 'dimensionless time'.
[Throughout] The paper does not provide a clear definition of the ensemble average notation; Fig. 2 uses <(X−Xbar)²> and <(C_F−C_Fbar)²> without specifying how the average over disorder realizations is taken or how many realizations are used. Reporting the number of realizations and error bars would strengthen the numerical claims.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the front-diffusion coefficient is derived analytically and then tested against independent numerics, and the genetic-loss claims are simulation-based with no fitted parameter renamed as a prediction.

full rationale

Walking the derivation chain, the effective front diffusion coefficient is obtained by a perturbative calculation, Eq. (2)-(11), that starts from the stochastic FKPP equation and the noise correlation <f(x)f(y)>=delta(x-y). The result D_C = (1/8) xi^2 sqrt(R D0) contains no parameter fitted from the data used for validation. The numerical test in Fig. 2 compares best-fit scaling exponents (xi^2.00+/-0.01, R^0.50+/-0.02, D0^0.50+/-0.02) against the derived formula, which is the correct, non-circular direction. For the genetic-loss part, the two-population model in Eq. (13) and the derived variance expression Eq. (19) are not used to select the nonmonotonic heterozygosity curve; the heterozygosity behavior in Fig. 3 comes from direct numerical solution of Eq. (13) across different xi values. There is no fitted parameter being fed back into a 'prediction' and no self-citation chain carrying the central claim; the cited references [5,20] supply a standard perturbation method, not an imported uniqueness result. One substantive correctness issue exists -- the homogeneous-environment baseline assertion that C1 'will be extinct, due to its smaller initial value' is doubtful because Eq. (14b) is a deterministic linear PDE for Y=C2-C1 with no stochastic source and a nonnegative growth term, so extinction is not an automatic consequence -- but this is a modeling/reference-behavior concern, not a circular reduction of the paper's derivation to its own inputs. Accordingly, the appropriate circularity score is 0.

Assumptions & free parameters 1 free parameters · 5 assumptions · 0 invented entities

The model introduces no new entities; the random diffusivity field f(x) is an environmental input, not a postulated new force or mediator. The most fragile premises are the inconsistent noise prescriptions and the unproven deterministic-extinction baseline.

free parameters (1)

xi_c threshold separating weak and strong disorder = approximately 0.2 (simulation-dependent; no analytic expression)
The boundary between weak and strong disorder in the genetic-loss claim is read off Fig. 3(b), not derived. It is part of the central qualitative claim but is not an input to the model.

assumptions (5)

domain assumption The Fisher-KPP reaction-diffusion equation with logistic growth describes invasion fronts and genetic competition.
Used as the model in Eq. (1) and Eq. (13).
ad hoc to paper The disorder function f(x) can be treated both as bounded uniform noise in [-1,1] and as delta-correlated Gaussian white noise with <f(x)f(y)> = delta(x-y).
The model paragraph defines f as uniform noise in [-1,1], the analytic calculation after Eq. (9) uses delta-correlation, and the simulation in Eq. (12) uses bounded disorder at grid scale. These prescriptions are not equivalent.
domain assumption At long times the front speed v equals 2 sqrt(R D0) even in the presence of weak spatial diffusivity disorder.
Invoked before Eq. (5) and used in the comoving frame perturbation; standard for FKPP but an approximation under disorder.
ad hoc to paper The product C times Upsilon squared can be neglected in Eq. (16) because C is near 1 where Upsilon is near 0 and C is near 0 where Upsilon is near 1.
Used to reduce the evolution equation for kappa; the overlap region at the front is assumed to contribute negligibly, which is not quantified.
ad hoc to paper In a homogeneous environment the initially less abundant population C1 goes extinct due to its smaller initial value.
Stated before Fig. 3 without derivation; Eq. (13) is deterministic and contains no demographic stochasticity, so this premise is not obviously valid.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Environmental Disorder Regulation of Invasion and Genetic Loss." pith.science (2026). https://pith.science/paper/IKZX7TEG

@misc{pith2026190802532,
  author       = {Pith},
  title        = {Pith review of: Environmental Disorder Regulation of Invasion and Genetic Loss},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/IKZX7TEG}},
  note         = {Machine review of arXiv:1908.02532}
}

read the original abstract

Many physical and natural systems, including the population of species, evolve in habitats with spatial stochastic variations of the individuals' motility. We study here the effect of those fluctuations on invasion and genetic loss. A Langevin equation for the \textit{position} and \textit{border} of the invasion front is obtained. A striking result is that small/large fluctuations of diffusivity suppress/intensify genetic loss. Our findings reveal the potential role of environmental fluctuations as a regulating factor for genetic loss and provide a simple explanation for the regional differences in the intensity of genetic drift observed during the final stages of human evolution and in tumor mutational landscapes.

Figures

Figures reproduced from arXiv: 1908.02532 by the authors.

**Figure 2.** FIG. 2: (a) [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 3.** FIG. 3: (a) Plots of the densities [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

28 extracted references · 27 canonical work pages

[1]

Kondo and T

S. Kondo and T. Miura, Science 329, 1616 (2010)

work page 2010
[2]

J. D. Murray, Mathematical Biology I: An Introduction (2003)

work page 2003
[3]

Hallatschek, Proceedings of the National Academy of Sciences 108, 1783 (2011)

O. Hallatschek, Proceedings of the National Academy of Sciences 108, 1783 (2011)

work page 2011
[4]

Hallatschek and K

O. Hallatschek and K. Korolev, Physical Review Letters 103, 108103 (2009)

work page 2009
[5]

Birzu, O

G. Birzu, O. Hallatschek, and K. S. Korolev, Proceedings of the National Academy of Sciences 115, E3645 (2018)

work page 2018
[6]

K. R. Swanson, C. Bridge, J. Murray, and E. C. Alvord Jr, Journal of The Neurological Sciences 216, 1 (2003)

work page 2003
[7]

Mandonnet, R

E. Mandonnet, R. Van Eﬀenterre, E. C. Alvord Jr, and L. Capelle, Annals of Neurology 53, 524 (2003)

work page 2003
[8]

Bordeleau et al., J

F. Bordeleau et al., J. Huynh, , Proceedings of the Na- tional Academy of Sciences 114, 492 (2017)

work page 2017

Show all 28 references

[9]

B. N. Mason, A. Starchenko, R. M. Williams, L. J. Bonas- sar, and C. A. Reinhart-King, Acta biomaterialia 9, 4635 (2013)

2013
[10]

Jamin, et al., Cancer Research (2015)

Y. Jamin, et al., Cancer Research (2015)

2015
[11]

Plodinec, et al., Nature Nanotechnology 7, 757 (2012)

M. Plodinec, et al., Nature Nanotechnology 7, 757 (2012)

2012
[12]

P. E. Howell, E. Muths, B. R. Hossack, B. H. Sigafus, and R. B. Chandler, Ecology 99, 1119 (2018)

2018
[13]

F. W. Allendorf, Zoo biology 5, 181 (1986)

1986
[14]

Hallatschek, P

O. Hallatschek, P. Hersen, S. Ramanathan, and D. R. Nelson, Proceedings of the National Academy of Sciences 104, 19926 (2007)

2007
[15]

Slatkin and L

M. Slatkin and L. Excoﬃer, Genetics 191, 171 (2012)

2012
[16]

Birzu, S

G. Birzu, S. Matin, O. Hallatschek, and K. S. Korolev, arXiv preprint arXiv:1903.11627 (2019)

2019 arXiv
[17]

Reiter, S

M. Reiter, S. Rulands, and E. Frey, Physical review let- ters 112, 148103 (2014)

2014
[18]

L. L. Cavalli-Sforza, P. Menozzi, and A. Piazza, Science 259, 639 (1993)

1993
[19]

Keinan, J

A. Keinan, J. C. Mullikin, N. Patterson, and D. Reich, Nature Genetics 39, 1251 (2007). 5

2007
[20]

Mikhailov, L

A. Mikhailov, L. Schimansky-Geier, and W. Ebeling, Physics Letters A 96, 453 (1983)

1983
[21]

Brunet and B

´E. Brunet and B. Derrida, Journal of Statistical Physics 103, 269 (2001)

2001
[22]

Sahimi, B

M. Sahimi, B. D. Hughes, L. Scriven, and H. T. Davis, Journal of Chemical Physics 78, 6849 (1983)

1983
[23]

M. J. Williams, B. Werner, C. P. Barnes, T. A. Graham, and A. Sottoriva, Nature genetics 48, 238 (2016)

2016
[24]

T. O. McDonald, S. Chakrabarti, and F. Michor, Nature genetics 50, 1620 (2018)

2018
[25]

Keinan, J

A. Keinan, J. C. Mullikin, N. Patterson, and D. Reich, Nature Genetics 41, 66 (2009)

2009
[26]

Fort and V

J. Fort and V. M´ endez, Physical Review Letters82, 867 (1999)

1999
[27]

M. O. Vlad and J. Ross, Physical Review E 66, 061908 (2002)

2002
[28]

J. Fort, J. P´ erez-Losada, and N. Isern, Physical Review E 76, 031913 (2007)

2007

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [1]

Kondo and T

S. Kondo and T. Miura, Science 329, 1616 (2010)

work page 2010

[2] [2]

J. D. Murray, Mathematical Biology I: An Introduction (2003)

work page 2003

[3] [3]

Hallatschek, Proceedings of the National Academy of Sciences 108, 1783 (2011)

O. Hallatschek, Proceedings of the National Academy of Sciences 108, 1783 (2011)

work page 2011

[4] [4]

Hallatschek and K

O. Hallatschek and K. Korolev, Physical Review Letters 103, 108103 (2009)

work page 2009

[5] [5]

Birzu, O

G. Birzu, O. Hallatschek, and K. S. Korolev, Proceedings of the National Academy of Sciences 115, E3645 (2018)

work page 2018

[6] [6]

K. R. Swanson, C. Bridge, J. Murray, and E. C. Alvord Jr, Journal of The Neurological Sciences 216, 1 (2003)

work page 2003

[7] [7]

Mandonnet, R

E. Mandonnet, R. Van Eﬀenterre, E. C. Alvord Jr, and L. Capelle, Annals of Neurology 53, 524 (2003)

work page 2003

[8] [8]

Bordeleau et al., J

F. Bordeleau et al., J. Huynh, , Proceedings of the Na- tional Academy of Sciences 114, 492 (2017)

work page 2017