Abstract
Endometriosis is associated with impaired endometrial decidualization, yet conventional differential expres-
sion analyses cannot readily distinguish delayed transcriptional activation from reduced response amplitude.
These mechanisms may produce similar expression differences at individual time points while implying
distinct biological interpretations. Establishing when they can be distinguished is therefore a computational
identifiability problem.
We implement an identifiability-aware analysis of temporal displacement and amplitude attenuation
in longitudinal transcriptomic responses. The approach combines reference trajectories, constrained
amplitude-only and amplitude-plus-delay models, analytical examples of observational equivalence, and
simulation-based assessments of mechanism distinguishability. Held-out time-point evaluation and culture-
level sensitivity analyses examine the stability of model-dependent conclusions.
In GSE246591, a candidate-delay-dependent dynamic-range filter excluded 30.9–38.3% of initially dynamic
features under a hypothetical 24-hour shift across four endometrial stromal cultures. A two-component
diagnostic separating retained dynamic range from affine shape similarity revealed that small projection
residuals can arise from either loss of observable dynamics or similarity of response shapes. These findings
demonstrate why a single separation statistic may be insufficient for interpreting temporal identifiability.
Complementary independent in vivo analyses provided biological context. In GSE51981, an endometrial
progesterone-response-associated transcriptional score was reduced in endometriosis samples (phase-model
estimate 0.155, 95% CI 0.222 to 0.088). GSE6364 showed a directionally consistent but statistically
inconclusive difference (0.033, 95% CI 0.068 to 0.002). Neither observation independently establishes a
temporal delay.
Together, these results support selection-aware evaluation of temporal transcriptomic responses, distin-
guishing structural ambiguity and loss of dynamic signal from evidence of a biological timing change.
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted November 4, 2026. ; https://doi.org/10.64898/2026.11.03.764027doi: bioRxiv preprint
1 Introduction
Decidualization is the hormone-associated differentiation of human endometrial stromal cells that prepares
the uterine lining for implantation [1]. An attenuated decidualization-associated transcriptional signal in
endometriosis has been reported in multiple endometrial studies [2, 3]. Yet a lower signal at one sampling time
can arise from a smaller response, a later response, altered trajectory shape, menstrual-cycle composition, or
an unmeasured cellular mixture. Distinguishing those explanations requires information beyond cross-sectional
differential expression. Menstrual-cycle-associated confounding further complicates endometrial comparisons
[4].
Temporal gene-expression methods answer related but distinct questions.DEtime estimates when two
perturbation-specific time courses diverge using Gaussian processes [5];ImpulseDE2 tests and models time-
dependent expression using an impulse family [6]. Neither method, without a suitable matched design and
specified nuisance parameters, establishes that a lower single-time signal is specifically a delayed version of
another response. Identifiability analysis asks whether different parameter settings can generate the same
observable data, a question with a long history in dynamical modeling [7, 8].
We operationalize the distinction in two steps. First, we identify when a shifted curve can be represented
exactly by changing only its amplitude and baseline (observational aliasing). Second, we quantify the
remaining separation on a verified endometrial RNA-sequencing schedule, with explicit checks of expression
assayability, reference-curve assumptions, and uncertainty limits. The resulting audit separates algebraic
nonidentifiability from what measured expression values can establish. A supplementary software example
documents a related computational workflow, but does not serve as biological validation.
2 What does it mean to identify a temporal delay?
2.1 A minimal model and an exact criterion
For sampling timesT ={t1,...,tm}and a specified reference expression profilef (t), compare an amplitude-
only response with an amplitude-plus-shift response:
M0 :y(t i) =b 0 +a 0f(ti) +εi,(1)
Mτ: y(ti) = bτ+aτf (ti−τ) +εi. (2)
Hereb is the baseline expression level,a is the response amplitude, andτis a displacement relative to a chosen
Reference
trajectory,not an experimentally observed molecular-onset time. Theεi represent measurement
error or departures from the assumed trajectory. We first consider the noise-free case with unconstrained
baseline and nonzero signed amplitudes.
Proposition 1(Observational equivalence on a finite time grid).Let f = (f (ti))m
i=1 and fτ= (f (ti−τ))m
i=1
.
The two response families can generate the same noiseless vector with nonzero amplitudes if and only if
fτ=cf+d1(3)
for an admissiblec̸= 0and offsetd. If amplitudes are restricted,c must also respect those restrictions.
Proof. Equality bτ1 +aτfτ= b01 +a0f yields c = a0/aτand d = (b0−bτ)/aτ. Conversely, substituting
Eq. (3) and settinga0 =aτc, b0 =bτ+aτd makes both vectors identical. Constant or zero-amplitude edge
cases must be treated separately.
For example,f (t) = ekt has f (t−τ) = e−kτf (t), whereasf (t) = αt+βhas f (t−τ) = f (t)−ατ. For a
fixed shift τ0, f (t) = ektq(t) with q(t−τ0) = q(t) also aliases. These are familiar algebraic counterexamples,
not new solution families. Such exact identities assume the shape extends over the shifted domain: a separate
rule that clampsf (t) before stimulation may break that identity.
1
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted November 4, 2026. ; https://doi.org/10.64898/2026.11.03.764027doi: bioRxiv preprint
2.2 A measurable, but conditional, diagnostic
Set X = [1, f ], letPX =XX + be the least-squares projector (withX + the Moore–Penrose pseudoinverse),
and define
δ(τ;T,f ) =∥(I−PX )fτ∥2, δ RMS =δ/√m.(4)
Proposition 2 (Signal–shape decomposition of a projection residual).Let g = fτbe the modeled shifted
curve and supposerange(f )> 0. DefineD =δRMS/range (f ), R = range(g)/ range(f ), and, for nonconstant
g,
S =∥(I−PX )g∥2
∥g−g1∥2
, D=R sd(g)
range(g)S≤R
2,(5)
where sd(g) uses divisorm. If g is constant,D = 0, R = 0, andS is undefined. For two varying curves, the
unconstrained signed-amplitude projection givesS2 = 1−r(f, g)2, wherer is their Pearson correlation.
Proof. Because 1∈col(X), the projection residual ofg is no larger than the centered norm ofg. With its
range B, sd(g)≤B/2 by the range–variance inequality. Substitution yields Eq.(5). The correlation identity
follows from ordinary least-squares regression of one centered vector on the other.
A smallD can therefore arise from weak retained signal (Rsmall) or similarity after affine reparameterization
(S small). These are different limits on interpretation. Neither quantity is a confidence interval or a calibrated
onset test. The two components do not alone determineD; its exact product also involvessd(g)/ range(g).
Which quantities remain identifiable? For the illustrative exponential referencef (t;k) = ekt, a delayed
response y(t) = b +aek(t−τ)can be written asb +Aekt with theidentifiable combinationA =ae−kτ. No
expression-only observation separatesa from τin this unrestricted family. Yet, in noise-free observations at
three equally spaced timest,t +h,t + 2h, a nonzero response withk̸= 0obeys
k =h−1 log
(︃y(t + 2h)−y(t +h)
y(t+h)−y(t)
)︃
.(6)
Thusk may remain identifiable even whenτdoes not. This is a model-specific illustrative reparameterization,
not a general estimator of a biological rate; near-flat curves or noisy observations can makek practically
unrecoverable [9].
Identifiability-aware feature eligibility.If Eq. (3) holds on the sampled grid, the identity
range(fτ) =|c|range(f ) (7)
is exact. Therefore requiringboth observed and modelled-shifted ranges to exceed a fixed threshold can exclude
some exactly aliased trajectories in a shift-dependent way. This conclusion concerns a candidate-dependent
shifted-range gate; it does not establish that conventional filtering on observed expression alone invalidates
temporal analyses. The direction of the selection effect depends on the trajectory, shift and pre-onset boundary
rule [10]. We report eligibility separately from projection residuals.
This is the residual after allowing the unshifted curve to change its baseline andsigned amplitude. Zero
residual denotes exact equivalence on the sampled grid; nonzero residual indicates geometric separation
conditional on the assumed reference and boundary rule. It is neither a p-value nor a confidence interval and
cannot determine an onset time by itself. With only two sampled times and unrestricted offset/amplitude,
even distinct shifts can always have zero residual whenX spans the observation space. Sampling more times
can break some ambiguities but cannot resolve a globally exponential or affine identity by itself.
3 Datasets and computational protocol
3.1 Observed endometrial time course
The Gene Expression Omnibus (GEO) datasetGSE246591 contains RNA-sequencing (RNA-seq) profiles of
four primary human endometrial stromal-cell cultures, S169, S170, S506, and S508, each sampled at0, 3, 6,
2
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted November 4, 2026. ; https://doi.org/10.64898/2026.11.03.764027doi: bioRxiv preprint
0 1 2 3 4 5 6 7 8
Illustrative time (arbitrary units)
1.0
1.5
2.0
2.5Reference response (arbitrary units)
Hypothetically shifted reference
Unshifted reference, rescaled
Illustrative sample times
Verify
GEO + 40 IDs
Normalize
log2(1+CPM)
Reference
Linear + pre-0
Compare
Shift vs scale
Report
Residual + limits
Shifted values are modeled from the same ten measured times; no observed biological-lag truth.
A Exact equivalence for an exponential reference
B Reproducible GSE246591 analysis
Figure 1: Exact ambiguity and analytical pipeline.(A) The two profiles overlap algebraically for
an illustrative exponential reference; dots arenot measured RNA-seq. (B) The mandatory audit sequence
checks sample identities, fixes the reference and nuisance parameters, compares amplitude with displacement,
assesses remaining geometric separation, and reports limits. The programmatic implementation records
source files and checksums.
9, 12, 18, 24, 36, 48, and 96 hoursafter decidualization induction [11]. We verified all 40 culture-by-hour
columns in the original deposited counts file (SHA-256 in Supplementary Methods). There are 60,676 source
feature rows. These are source identifiers, not necessarily 60,676 uniquely mapped genes. No independent
within-culture, same-hour replicate series is available for estimating measurement variance or ground-truth
delay.
For each feature and culture, raw counts were divided by the total reads assigned to the sample, scaled
to one million (counts per million; CPM), and transformed aslog2(1 + CPM). We considered hypothetical
displacements τ∈{3, 6, 12, 24}hours. The measured ten-point curve was connected bypiecewise-linear
interpolation; for shifted arguments below time zero we imposedf (t < 0) = f (0). No expression at an
unmeasured shifted time was treated as a new observation. For each candidate displacement we computed
Eq. (4) and dividedδRMS by the measured ten-point expression range. Features entered each descriptive
summary only if both their measured and modeled shifted expression ranges were at least 0.5 log2(1+CPM).
Consequently, the set of included features changes with shift and culture; our summaries are not paired
gene-level contrasts.
3.2 Evidence roles and safeguards
Sample identities, input checksums, eligibility rules and model-specific statistics are recorded in the Supplement.
We present no estimate of power or classification accuracy for molecular onset. The threshold of normalized
3
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted November 4, 2026. ; https://doi.org/10.64898/2026.11.03.764027doi: bioRxiv preprint
Table1: Distinctanalyticalroles. Thereporteddatasetsare not pooled; samplesdonotrepresentexchangeable
units across studies.
GEO series Design Role / boundary
GSE246591 Four cultures;
ten sampling
times each
Reference-shift geometry; no known biological lag or within-time replicate
variance.
GSE94036 PGR/FOXO1
perturbation
in stromal
cells
Source of the PGR-oriented score, not a timed-lag experiment.
GSE51981 Tissue
microarrays;
MSE subset
28 vs 8
Fixed-score clinical contrast, not measured delay.
GSE6364 Tissue
microarrays;
MSE subset 9
vs 8
External directional score check, not an hourly time course.
projection residual below 0.05 isdescriptive only: it is not derived from the experimental error distribution. We
separately describe a retrospective shape-sensitivity exercise on 48 features but do not call it a head-to-head
benchmark with DEtime or ImpulseDE2.
4 Results
4.1 Exact counterexamples: why timing alone may be unidentifiable
All three algebraic examples in Fig. 2 satisfy Eq.(3) for the specified shifts and allowed nuisance parameters.
Thus zero-noise expression values alone need not distinguish a temporal shift from rescaling or changing
Background
expression. These mathematical witnesses do not establish that the corresponding shape is
biologically correct. A separate historical implementation verified the three identities to numerical tolerance;
we treat the identities themselves, rather than an inferred biological lag, as the result.
4.2 Real measured trajectories show conditional, heterogeneous separation
The deposited GSE246591 counts matrix passed the 40/40 sample-identity and integer-count checks. Across
the four cultures and four shifts we obtained 16 descriptive summaries (Supplementary Table S1). Forτ= 3
hours, medianδRMS/range(f ) was 0.220 for S169 and 0.153–0.165 for the remaining cultures. At 6 hours
the four medians were 0.206–0.214; at 12 hours they were 0.226–0.231; at 24 hours they were 0.209–0.219
(Fig. 3). Thus a general assertion thatmost measured response curves are nearly exactly compatible with a
shifted-and-rescaled response is not supported by this diagnostic.
At the descriptive cutoffδRMS/range(f )< 0.05, the fraction of included features ranged from approximately
0.02% to 3.20% across the 16 culture-by-shift settings. This is neither an estimate of the proportion of
genuinely delayed genes nor a noise-calibrated test: the source includes only one profile per culture-hour
and the hypothetical shifted values are computed from the same measured series. Larger shifts also change
the eligible feature universe; for example S169 included 14,019 rows at three hours but 9,116 at 24 hours.
We therefore do not interpret across-shift median differences as paired gene-level effects. The reported
proportions also condition onboth dynamic-range gates. In particular, exact aliased profiles with shrinking
modeled response ranges need not remain eligible (Eq.(7)). A selection-aware audit instead reports the
denominator passing the observed-range gate, the additional exclusions caused by the candidate shift, and
4
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted November 4, 2026. ; https://doi.org/10.64898/2026.11.03.764027doi: bioRxiv preprint
0.0 2.5 5.0 7.5 10.0
Time (illustrative)
0.2
0.4
0.6
0.8
1.0Normalized responseshift = scale
A Exponential
Shifted
Affine-reexpressed
0.0 2.5 5.0 7.5 10.0
Time (illustrative)
0.2
0.4
0.6
0.8
1.0
shift = baseline change
B Affine
0.0 2.5 5.0 7.5 10.0
Time (illustrative)
0.2
0.4
0.6
0.8
1.0
at shift = 6
C Exponential × periodic
Figure 2: Exact analytic counterexamples, not measured data.An exponential, an affine function
and an exponential multiplied by a periodic factor can each alias a specified shift through scale and/or offset.
The periodic construction establishes equality for its stated shift, not all possible shifts. The curves here are
unclamped; the empirical analysis uses an explicit flat pre-zero boundary.
the distribution of the same geometric residual in retained and excluded features (Supplementary Methods).
This is a retrospective diagnostic, not a population estimate of lag identifiability.
4.3 Separating signal loss from affine shape similarity
The two-component audit of the same GSE246591 matrix identifies distinct explanations for a smallD
(Fig. 4). Among features passing the initial observed-curve range screen, 982 culture–feature records at
a three-hour candidate shift hadD < 0.05; 977 (99.5%) also hadS < 0.20, whereas none hadR <0.20.
Within each culture the near-zero group’s medianR was 0.968–0.975 and medianS was 0.131–0.141. Thus
short-shift small residuals commonly arose from affine shape similarity despite retained dynamic range, under
the specified reference model.
At a 24-hour shift, 730 culture–feature records were both excluded by the shifted-range screen and below
D <0.05. Of these, 557 (76.3%) hadR< 0.20, while only 20 (2.7%) hadS <0.20; one shifted curve was
exactly flat and had undefinedS. MedianR among these low-D excluded features was 0.136–0.186 across
cultures, whereas median definedS was 0.866–0.989. Their small residuals more often reflected loss of the
shifted signal than similarity of two varying shapes. The 122 near-zero records that remained shift-eligible at
24 hours showed comparatively higher medianR (0.578–0.616) and lowerS (0.225–0.236). The illustrative
0.20cutoffs do not define statistical significance.
The distinction also explains why the secondary dynamic-range gate changes the denominator: of 56,988
initially dynamic culture–feature records at 24 hours, 19,676 (34.5%) were excluded by the shifted-range
condition. Among all initially eligible records, 852 (1.50%) hadD< 0.05, compared with 122/37,312 (0.33%)
of the retained set. We donot interpret this difference as proof that filtering preferentially removes structurally
aliased trajectories: Eq.(5) makes smallD mechanically possible whenR is small. A fixed intersection of
eligible features across all four candidate shifts (7,858–9,960 rows per culture) provides a stable denominator
for supplementary descriptive comparisons. Additional culture- and shift-specific summaries are reported in
the Supplement.
4.4 Sensitivity to the assumed reference-curve shape
A separate, retrospectively selectedfixed 48-feature subset was evaluated using piecewise-linear interpolation,
a Gaussian-process (GP) approximation with fixed parameters, and a two-sigmoid impulse-like approximation.
In leave-one-time-point-out reconstruction, linear interpolation had the lowest median feature-level root-mean-
square error (RMSE) in each culture (0.595–0.737 log2CPM). The GP and impulse-like curves often produced
smaller conditional residuals for a three-hour hypothetical shift, illustrating the sensitivity of Eq.(4) to the
5
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted November 4, 2026. ; https://doi.org/10.64898/2026.11.03.764027doi: bioRxiv preprint
3 6 12 24
Hypothetical displacement τ (hours)
0.12
0.14
0.16
0.18
0.20
0.22
0.24
0.26Median projected residual / observed range
S169 S170 S506 S508
Figure 3: Measured-source, model-conditional GSE246591 geometry.Medians over included feature
rows of the root-mean-square residual after projecting a hypothetical displaced curve onto baseline and
signed amplitude, divided by each feature’s measured expression range. The values are computed from actual
GEO counts, while displaced coordinates rely on piecewise-linear interpolation and the flat pre-zero rule.
Feature eligibility differs byτ(Supplementary Fig. S3); connecting lines is for readability, not evidence of a
continuously estimated lag.
Reference
representation. The pilot emitted numerical warnings from matrix operations despite finite exported
tables. We therefore classify this exercise asexploratory and not numerical method validation; its complete
descriptive readouts and caveats are in Supplementary Table S3. These approximations are not the published
implementations of DEtime or ImpulseDE2; they serve only to test sensitivity to curve representation.
4.5 Independent biological context does not establish temporal delay
The PGR (progesterone receptor)-dependency score was derived from a separate stromal-cell perturbation
experiment (GSE94036) and evaluated in endometrial cohorts [2, 3, 12–14]. In the 28 endometriosis versus
eight strict-control mid-secretory endometrium (MSE) samples from GSE51981, the estimated score difference
was−0.154922. The phase-interaction modelusing all eligible menstrual phases (n = 109) yielded a 95%
confidence interval (CI) based on HC3 heteroskedasticity-robust standard errors[−0.222227,−0.087617] for
the MSE contrast andp = 1.386×10−5. A separate mid-secretory-only model with HC3 standard errors
produced [−0.223889,−0.085955] (p = 6.247×10−5); adjusting that MSE-only model for three composition-
proxy components gave−0.076561 (95% CI [−0.133443,−0.019678], p = 0.009975). These are different
models and must not be assigned one another’s p-values.
In the independent GSE6364 mid-secretory subset (nine endometriosis cases and eight controls), the
frozen-score difference was−0.032765. Its unequal-varianceWelch 95% CI was[−0.067656, +0.002126], with
p = 0.063752. Its direction was consistent but its interval included zero. The GSE51981 values from the phase
model and the GSE6364 Welch estimate explain the rounding in the registered abstract. Neither endometrial
comparison repeatedly samples thesame women over hours after a controlled perturbation; no delayed
molecular onset can be inferred from them. We do not pool effect sizes across these distinct estimands.
6
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted November 4, 2026. ; https://doi.org/10.64898/2026.11.03.764027doi: bioRxiv preprint
0.0 0.2 0.4 0.6 0.8 1.0
Dynamic-range retention, R
0.0
0.2
0.4
0.6
0.8
1.0Affine shape separation, S
Four cultures per stratum
A Near-zero residual strata (culture medians)
3 h: initially dynamic
24 h: shift-excluded
24 h: shift-retained
3 h: initial dynamic
(n=982)
24 h: shift-excluded
(n=730)
0
20
40
60
80
100Within near-zero residuals (%)0.0%
76.3%
99.5%
2.7%
B Descriptive component thresholds
R < 0.20
S < 0.20
GSE246591 · Median points and cutoff counts from author-verified summary only; not per-gene scatter or calibrated rates.
Figure 4: Two explanations for a small geometric projection residual. (A) Each point is a
culture-level median among near-zero(D <0.05) records, not a per-gene joint scatter; the three symbols
distinguish the three-hour initially dynamic stratum from 24-hour shifted-range-excluded and retained strata.
(B) Descriptive percentages belowR = 0.20 and S = 0.20 within the corresponding near-zero groups. The
groups differ in size and selection rule; thresholds are illustrative, not calibrated tests. Source: GSE246591
and the archived author-run dual-component audit.
5 Discussion
The central distinction is betweenstructural ambiguity(different mechanisms giving exactly the same noiseless
observations) andempirical information (the separation suggested by an observed curve under specified
assumptions). Algebra shows that some shape families can alias delay exactly. With an empirical curve,
projection residuals quantify the distance between hypothetical shifted and unshifted profiles only after
Reference
interpolation, boundary behavior, background and amplitude freedom have been specified. Even
a nonzero residual cannot imply practical distinguishability without replicate-level noise and appropriate
dependence modeling [
7, 8]. The actual GSE246591 grid showed heterogeneous, mostly non-negligible
normalized distances among the eligible features rather than pervasive near-zero residuals. That descriptive
finding refines the motivating question; it does not recover the onset of a biological event.
A constructive positive implication is that when a shift is absorbed by an amplitude–baseline transformation,
a parameter describingshape may remain measurable. For an exponential family, the ratek is an example
(Eq. (6)); time to a specified observed plateau is not automatically an identifiable substitute and would
require its own model-specific analysis. The appropriate estimand therefore depends on which parameters
survive the nuisance reparameterization.
The signal–shape decomposition additionally shows why a low range-normalized residual is not, by itself,
evidence of affine equivalence: reduced shifted dynamic range can lower the residual without reducing the
relative shape-separation ratio. We recommend reporting signal retention, affine shape separation, non-
evaluable flat profiles and eligibility counts together, rather than using a single near-zero residual fraction.
This is an interpretive auditing protocol built from standard regression geometry, not a novel statistical test
or a proven method for recovering onset.
The main limits are a single measured expression vector per culture-hour, no known imposed lag, no
matched untreated curve at every time, uncertain interpolation and pre-zero behavior, source feature identifiers
not mapped to unique genes, and retrospective eligibility/filter choices. Thousands of feature rows should not
be counted as thousands of independent biological replicates. In addition, the amplitude projection is signed;
a constrained nonnegative-amplitude model would yield different distances when fitted slopes are negative.
The Supplement distinguishes that issue and the separate fixed-panel historical analyses. The curve-shape
pilot is not a validated comparison with published state-of-the-art methods and emitted numerical warnings;
its values must not be used to rank algorithms.
Related preprints by the present authors discuss assayability and cohort transfer [15], and the distinction
between computational stability and biological interpretation [16]. They motivate separate reporting of
7
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted November 4, 2026. ; https://doi.org/10.64898/2026.11.03.764027doi: bioRxiv preprint
measurement compatibility, statistical reproducibility and biological meaning, but do not validate this
endometrial timing hypothesis. A decisive follow-up would preregister amplitude and curve-shape constraints,
collect matched untreated time courses from independent donors with within-time biological replicates, impose
a known perturbation delay, and assess candidate estimators on the same prediction target. Such a study
would also permit a fair comparison with DEtime, ImpulseDE2 and other established time-course approaches.
Where those conditions are absent, the correct result is a limit on interpretability, not a claimed delayed
response.
6 Conclusions and software availability
Delay and attenuation can be observationally equivalent even with perfect measurements, but how closely they
agree on a real sampling grid is reference- and assumption-dependent. Our endometrial RNA-seq audit provides
reproducible descriptive evidence for that distinction without claiming biological-onset recovery or a new
identification theorem. Code and reproducibility instructions: https://github.com/Helen2813/paper7.
The separate OnsetAudit application, its genuine screenshot and its mouse-cell elapsed-exposure-time
benchmark are documented in the Supplement; they are not evidence for human endometrial lag.
8
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted November 4, 2026. ; https://doi.org/10.64898/2026.11.03.764027doi: bioRxiv preprint
References
[1] Birgit Gellersen and Jan J. Brosens. Cyclic decidualization of the human endometrium in reproductive
health and failure.Endocrine Reviews, 35(6):851–905, 2014. doi: 10.1210/er.2014-1045.
[2] Richard O. Burney, Said Talbi, Amy E. Hamilton, Kim Chi Vo, Mette Nyegaard, Camran R. Nezhat,
Bruce A. Lessey, and Linda C. Giudice. Gene expression analysis of endometrium reveals progesterone
resistance and candidate susceptibility genes in women with endometriosis.Endocrinology, 148(8):
3814–3826, 2007. doi: 10.1210/en.2006-1692.
[3] John S. Tamaresis, Juan C. Irwin, Gabriel A. Goldfien, Joseph T. Rabban, Richard O. Burney, Camran
Nezhat, Louis V. DePaolo, and Linda C. Giudice. Molecular classification of endometriosis and disease
stage using high-dimensional genomic data.Endocrinology, 155(12):4986–4999, 2014. doi: 10.1210/en.
2014-1490.
[4] Almudena Devesa-Peiro, Patricia Sebastian-Leon, Antonio Pellicer, and Patricia Diaz-Gimeno. Guidelines
for biomarker discovery in endometrium: Correcting for menstrual cycle bias reveals new genes associated
with uterine disorders. Molecular Human Reproduction, 27(4):gaab011, 2021. doi: 10.1093/molehr/
gaab011.
[5] JingYang, ChristopherA.Penfold, MurrayR.Grant, andMagnusRattray. Inferringtheperturbationtime
from biological time course data.Bioinformatics, 32(19):2956–2964, 2016. doi: 10.1093/bioinformatics/
btw329.
[6] David S. Fischer, Fabian J. Theis, and Nir Yosef. Impulse model-based differential expression analysis of
time course sequencing data.Nucleic Acids Research, 46(20):e119, 2018. doi: 10.1093/nar/gky675.
[7] Andreas Raue, Clemens Kreutz, Tim Maiwald, Julie Bachmann, Marcel Schilling, Ursula Klingmueller,
and Jens Timmer. Structural and practical identifiability analysis of partially observed dynamical models
by exploiting the profile likelihood.Bioinformatics, 25(15):1923–1929, 2009. doi: 10.1093/bioinformatics/
btp358.
[8] Franz-Georg Wieland, Adrian L. Hauber, Marcus Rosenblatt, Christian Tönsing, and Jens Timmer.
On structural and practical identifiability.Current Opinion in Systems Biology, 25:60–69, 2021. doi:
10.1016/j.coisb.2021.03.005.
[9] Nicolette Meshkat, Marisa Eisenberg, and Joseph J. DiStefano. An algorithm for finding globally
identifiable parameter combinations of nonlinear ODE models using Gr"obner bases.Mathematical
Biosciences, 222(2):61–72, 2009. doi: 10.1016/j.mbs.2009.08.010.
[10] Richard Bourgon, Robert Gentleman, and Wolfgang Huber. Independent filtering increases detection
power for high-throughput experiments. Proceedings of the National Academy of Sciences, 107(21):
9546–9551, 2010. doi: 10.1073/pnas.0914005107.
[11] NCBI Gene Expression Omnibus. GSE246591: Transcriptomic exploration of decidualizing human
endometrial stromal cell dynamics. Gene Expression Omnibus dataset, 2024. URLhttps://www.ncbi.
nlm.nih.gov/geo/query/acc.cgi?acc=GSE246591.
[12] NCBI Gene Expression Omnibus. GSE94036: Progesterone receptor- and FOXO1-dependent transcrip-
tomes of decidualized human endometrial stromal cells. Gene Expression Omnibus dataset, 2017. URL
https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE94036.
[13] NCBI Gene Expression Omnibus. GSE51981: Molecular classification of endometriosis and disease stage.
Gene Expression Omnibus dataset, 2014. URL
https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?
acc=GSE51981.
[14] NCBI Gene Expression Omnibus. GSE6364: Gene profiling of endometrium reveals progesterone
resistance in endometriosis. Gene Expression Omnibus dataset, 2007. URL
https://www.ncbi.nlm.
nih.gov/geo/query/acc.cgi?acc=GSE6364.
[15] Elena Krikun and Abedalrhman Alkhateeb. External evaluation of multi-omics prognostic models across
TCGA-BRCA and METABRIC: Transportability, stability, and incremental performance. bioRxiv, 2026.
URL https://doi.org/10.64898/2026.09.18.752709. Preprint; not peer reviewed.
[16] Elena Krikun and Abedalrhman Alkhateeb. Latent-factor stability is not sufficient: Evaluating SABLE-
MO in simulated and cancer multi-omics data. bioRxiv, 2026. URLhttps://doi.org/10.64898/2026.
10.12.758631. Preprint; not peer reviewed.
9
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted November 4, 2026. ; https://doi.org/10.64898/2026.11.03.764027doi: bioRxiv preprint
Text is read by the "Ask this paper" AI Q&A widget below.
Extraction quality varies by source — PMC NXML preserves structure
cleanly, OA-HTML may include some navigation residue, and OA-PDF can
have broken hyphenation. The publisher copy
(via DOI)
is the canonical version.