Pith. sign in

REVIEW 3 major objections 7 minor 88 references

Global testing of SNP-methylation interactions on binary phenotypes via a logistic functional regression model

T0 review · 3 major / 7 minor · reviewed 2026-07-13 · grok-4.5

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

Pith's one-line read A global likelihood-ratio test finds regional SNP–methylation interactions on binary disease risk by treating methylation as a smooth curve weighted near each SNP.

desk verdict Solid, usable extension of the authors’ continuous-outcome SNP–methylation interaction model to logistic outcomes; calibrated in sims and more powerful than pairwise screens when signal is regional, with the usual unproved Laplace LRT pivot as the main soft spot. read the letter →

arxiv 2607.09535 v1 pith:DEPTRW5F submitted 2026-07-10 stat.ME

Yvelin Gansou , Karim Oualkacha , Marzia Angela Cremona , Lajmi Lakhal-Chaieb This is my paper
classification stat.ME MSC 62J1262P1062G08
keywords SNP-methylationinteractionlogisticfunctionalregressionDNAmethylationbinaryphenotypegloballikelihood-ratiotestlocalizedkernelpenalizedGLMMregion-levelinference
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

Binary outcomes such as disease status are hard to link to both genetics and DNA methylation at once: classical logistic regressions on every SNP–CpG pair explode into thousands of unstable tests and ignore the fact that neighbouring methylation sites move together. This paper replaces those pairwise tests with a single logistic functional regression in which each person’s methylation measurements are first smoothed into a continuous curve along the genome; each SNP then multiplies a local kernel-weighted integral of that curve, so interaction strength decays with genomic distance from the SNP. Estimation is performed by penalized iteratively reweighted least squares inside a generalized linear mixed model, and a Laplace-approximated likelihood-ratio statistic supplies one global p-value for the whole set of interactions. Simulations show the test keeps its nominal size and gains power once the true signal spans multiple nearby CpGs; an obesity case–control application detects a significant regional interaction and improves classification over the no-interaction model. The result matters because it turns an intractable multiple-testing problem into a stable, biologically local region-level statement about how genotype and epigenotype jointly shape binary risk.

What carries the argument

The localized interaction term φ_d(t,u_d)=η_d ψ_ρ(|t-u_d|) that multiplies genotype by a distance-decaying kernel of the methylation curve; after B-spline expansion this produces a finite-dimensional logistic mixed model whose Laplace-approximated likelihood-ratio statistic tests H0: all η_d=0.

What would settle it

Generate binary data under a pure main-effects logistic model (no SNP–methylation interaction) at sample sizes matching the paper’s simulations; if the reported p-value distribution systematically exceeds the uniform diagonal on QQ plots, the claimed calibration fails.

Watch

Extended reading notes

Core claim

A logistic functional regression that lets each SNP interact with a kernel-localized integral of a smooth methylation curve yields a single, well-calibrated likelihood-ratio test for the overall presence of SNP–methylation interaction on a binary phenotype; the test controls type I error and is more powerful than classical SNP–CpG pairwise logistic analyses when the interaction signal is regional.

Load-bearing premise

The Laplace-approximated likelihood-ratio statistic is treated as asymptotically chi-squared with degrees of freedom equal to the number of SNPs, even though the smoothing parameter is estimated and the kernel scale is user-chosen.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 7 minor

Summary. The paper develops a logistic functional regression model for global testing of SNP–methylation interactions on binary phenotypes. Methylation at CpG sites is converted to smooth curves that interact with discrete SNP genotypes through a localized distance kernel ψ_ρ; the functional main effect is expanded in B-splines and estimated by penalized likelihood in a GLMM formulation (PIRLS with Laplace-approximated REML for the smoothing parameter). A likelihood-ratio test of the joint interaction coefficients η is proposed and claimed to be asymptotically χ²_D under H0. Simulations (null QQ plots with 5000 replicates; power under ρ misspecification; real-data-based power vs pairwise logistic tests) and an application to the GSE73103 obesity case–control data (region-level p = 0.043; improved AUC) are used to support calibration, power gains when interaction signal is regional, and practical utility.

Significance. If the calibration and power claims hold, the work fills a clear methodological gap: region-level SNP–methylation interaction testing for case–control outcomes that borrows strength across spatially correlated CpGs and avoids the multiple-testing burden of SNP–CpG pairwise logistic analyses. The binary extension of the authors’ continuous-outcome framework is nontrivial (nonlinear link, separation risk, Laplace REML). Concrete strengths include extensive null simulations across kernels and ρ, robustness checks for moderate ρ misspecification, a public-data illustration with improved discrimination, and a documented R package (funInterMethSNP) covering continuous, binary, and count outcomes. The method is best suited to targeted or region-based analyses of a moderate number of SNPs rather than genome-wide scans, which the authors acknowledge.

major comments (3)
  1. §2.5 asserts that the Laplace-approximated marginal LRT Λ is asymptotically χ²_D under H0 (D = number of SNPs), with no derivation or citation covering the combination of a nonlinear logit link, outer REML estimation of λ, and a user-fixed interaction kernel scale ρ that defines the design columns K_ρ. All reported p-values (simulations and the real-data p = 0.043) rest on this pivot. The QQ plots in Figs. 1–3 (N = 400/800, 5000 reps) are reassuring for the settings tried, but they do not replace a theoretical or resampling justification for this GLMM. The manuscript should either (i) supply references/theory for fixed-effect LRTs under Laplace-approximated logistic GLMMs with estimated smoothing parameters, or (ii) implement and recommend a parametric bootstrap (or other finite-sample calibration) as the primary or default inferential procedure, and state clearly when the χ²_D approxima
  2. The interaction design depends on a free scale ρ that is treated as known when forming A_ρ and when computing Λ (§2.2–2.5). Power under misspecified ρ is studied (§3.2, Figs. 4–5), but type I error under kernel/ρ misspecification is only checked under the pure null (η = 0), and the real-data LRT is reported only for ρ = 10 (§4.2). Because different ρ values change the effective interaction columns, reporting a single p-value without a selection rule or multiplicity adjustment over ρ is incomplete. Provide practical guidance for choosing ρ (or a small grid), show null calibration when ρ is selected from data, and either pre-specify ρ or account for searching over ρ when claiming significance.
  3. Introduction and §2.4 correctly flag separation and unstable interaction estimates as central difficulties for logistic models, yet the estimation and testing sections do not describe diagnostics or safeguards when SNP–methylation products nearly separate cases and controls (especially with low-MAF SNPs and localized kernels). PIRLS with a roughness penalty on δ(t) alone does not automatically regularize η. Clarify finite-sample behavior under quasi-complete separation, and state whether additional penalties, Firth-type adjustments, or constraints on η are used or recommended; otherwise the claimed stability advantage over pairwise LogittestSNPCpG remains only partially supported.
minor comments (7)
  1. §5: typo “developped” → “developed”.
  2. Notation: μ_i is defined as P(Y_i = 1 | ·) but Y_i ∈ {0,1} is written without braces in places (§2.1); keep set notation consistent.
  3. Figures 1–3: axis labels use mixed en-dashes and hyphens; ensure consistent −log10 notation and that the 45° reference line is stated in the captions.
  4. §4.1: the genomic gap and partition into region 1 vs 2 are clear, but a one-sentence justification for excluding region 2 beyond “sparse coverage” would help reproducibility.
  5. §3: power curves under the alternative are shown only for N = 400; a brief N = 800 power panel (or statement) would align with the null design.
  6. References: ensure Gansou et al. (2026) citation details match the published Statistics in Medicine entry once finalized; arXiv version numbering should be consistent.
  7. Package URL is given; a short note on version pinning or vignette reproducing Figs. 1–3 / Fig. 7 would strengthen reproducibility claims.

Circularity Check

2 steps flagged · score 2.0 of 10

Minor self-citation imports the continuous-outcome localized kernel and methylation smoothing; the logistic GLMM, PIRLS estimation, Laplace LRT and all binary simulations are independently specified.

  1. ansatz smuggled in via citation [Section 2.2, definition of φ_d and Eq. (3)]
    "To reflect the intuition that interactions are stronger near the SNP position ud, following Gansou et al. (2026) we define a localized bivariate interaction term: φd(t,ud)=ηdψρ(|t−ud|), where ηd is a scalar regression coefficient and ψρ:[0,1]→[0,1] is a decreasing kernel function parameterized by a positive parameter ρ… We consider three shapes for ψρ, namely: convex ψρ(u)=exp(−ρu), concave ψρ(u)=exp(−ρ2u2) and linear ψρ(u)=max(1−ρu,0)."

    The spatial interaction structure that the global test is designed to detect is adopted wholesale by citation to the authors’ own continuous-outcome paper rather than re-derived for the binary setting. While the subsequent logistic estimation and LRT are new, the load-bearing modeling ansatz (localized kernel weighting by genomic distance) is justified only by that self-citation.

  2. self citation load bearing [Section 2.1, functional representation of methylation]
    "Full details of the smoothing procedure, including the adaptive bandwidth selection, are provided in Gansou et al. (2026)."

    Construction of the functional predictors Π_i(t) that enter every subsequent integral and design matrix is deferred entirely to the authors’ prior paper; the present work supplies no independent derivation or external reference for the smoother that underpins the whole analysis.

full rationale

The paper’s central claim is a well-calibrated global LRT for SNP–methylation interactions under a logistic functional model. The binary likelihood (Eq. 6), PIRLS updates, Laplace-approximated REML criterion Vr(λ), and the χ²_D LRT (Section 2.5) are written out from first principles for the Bernoulli–logit GLMM and are checked by 5000 null simulations (Figs. 1–3) plus power curves against an external pairwise competitor. The only self-referential elements are the functional representation of methylation (kernel smoother, adaptive bandwidth) and the distance-based interaction kernel φ_d(t,u_d)=η_d ψ_ρ(|t-u_d|), both taken from the authors’ own continuous-outcome precursor Gansou et al. (2026). These are modeling choices, not fitted constants that later reappear as “predictions,” and the binary inferential machinery does not reduce to them by construction. ρ remains a user-chosen sensitivity parameter whose misspecification is explicitly stress-tested. No uniqueness theorem, no fitted-input-called-prediction, and no algebraic identity equating output to input appear. Hence only a low circularity score is warranted.

Assumptions & free parameters 5 free parameters · 6 assumptions · 2 invented entities

The central claim rests on standard GLM/GLMM math plus domain modeling choices for methylation smoothness and distance-decaying SNP–methylation interaction. Free parameters (ρ, basis size, smoothing bandwidth, estimated λ) and the unproven χ² pivot for the Laplace LRT are the main non-free lunch items; no new physical entity is postulated.

free parameters (5)
  • ρ (interaction kernel scale)
    User-chosen decay of ψ_ρ; not estimated from data. All real-data and many simulation fits fix ρ (e.g. ρ=10). Results are reported as moderately robust but still depend on this choice.
  • L (number of B-spline basis functions for δ(t))
    Fixed at L=10 in simulations and application; controls flexibility of the main methylation effect.
  • λ (roughness penalty / smoothing parameter)
    Estimated by Laplace-approximated REML-type criterion; free in the statistical sense and enters both estimation and the LRT comparison.
  • Methylation smoother bandwidth (k, h_min)
    Adaptive Gaussian kernel uses k=70 and h_min=1 kb following BSmooth defaults; shapes the functional predictors Π_i(t) before modeling.
  • γ (simulation interaction strength modulator)
    Scales α and η in data generation; used to draw power curves, not estimated on real data, but drives the reported power claims.
assumptions (6)
  • domain assumption Conditional phenotype follows a Bernoulli logistic model with linear predictor including main SNP, functional methylation, and kernel-weighted interactions.
    Equation (2)–(3); standard GLM assumption for binary traits, not verified beyond simulation design.
  • domain assumption Discrete CpG methylation can be treated as noisy samples of a smooth function Π_i(t) on the genomic domain.
    Section 2.1 and equation (1); borrowed from functional epigenomics literature and the authors’ prior work.
  • ad hoc to paper SNP–methylation interaction intensity decays with genomic distance via ψ_ρ with ψ_ρ(0)=1 (convex, Gaussian, or linear forms).
    Section 2.2; modeling device introduced in Gansou et al. (2026) and reused here without independent biological identification of the true decay shape.
  • standard math Spline coefficients b are Gaussian random effects with precision λP_1p, justifying GLMM estimation and Laplace marginal likelihood.
    Section 2.4; standard roughness-penalty / mixed-model equivalence (Wahba, Wood).
  • ad hoc to paper Under H0: η=0, the Laplace-approximated LRT is asymptotically χ²_D.
    Section 2.5; stated without proof for this estimated-λ logistic functional setting; calibration checked only by simulation QQ plots.
  • domain assumption Individuals are independent; sample size is large enough relative to the number of fixed effects including D interaction coefficients.
    Section 2 and Discussion; authors themselves note instability when D is large relative to N.
invented entities (2)
  • Localized bivariate interaction coefficient φ_d(t,u_d)=η_d ψ_ρ(|t−u_d|)
    purpose: Encodes distance-weighted SNP×methylation interaction inside the logistic functional model so a single η_d captures regional interaction per SNP.
    Modeling construct carried from the continuous-outcome predecessor; no independent measurement of φ_d outside the fitted model.
  • Global LRT Λ for joint SNP–methylation interaction on binary outcomes under the Laplace GLMM likelihood
    purpose: Provide one region-level test instead of many pairwise SNP–CpG tests.
    The test statistic is defined from the paper’s approximated marginal likelihoods; validity rests on the χ²_D axiom above.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Global testing of SNP-methylation interactions on binary phenotypes via a logistic functional regression model." pith.science (2026). https://pith.science/paper/DEPTRW5F

@misc{pith2026260709535,
  author       = {Pith},
  title        = {Pith review of: Global testing of SNP-methylation interactions on binary phenotypes via a logistic functional regression model},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DEPTRW5F}},
  note         = {Machine review of arXiv:2607.09535}
}
read the original abstract

Understanding how genetic and epigenetic factors jointly influence binary health outcomes remains a major challenge in biomedical research. We propose a global test for the overall effect of interactions between DNA methylation and a set of single nucleotide polymorphisms (SNPs) on a binary phenotype. We propose a logistic functional regression model in which methylation measurements at CpG sites are transformed into smooth functional predictors interacting with discrete SNP genotypes through a localized kernel. This framework enables stable inference on region-level interactions while accounting for the spatial structure of methylation around SNPs. Extensive simulations show that the proposed test provides well-calibrated type I error and improved power over classical SNP-CpG pairwise analyses. The practical relevance of the method is illustrated using publicly available methylation and genotyping data from an obesity case-control study.

Figures

Figures reproduced from arXiv: 2607.09535 by the authors.

Figure 1
Figure 1. Quantile–quantile plots of p-values from 5000 simulations under the null hypothesis (H0). The model is fitted using the convex kernel ψρ(u) = e −ρu . 12 [PITH_FULL_IMAGE:figures/full_fig_p012_1.png] view at source ↗
Figure 2
Figure 2. Quantile–quantile plots of p-values from 5000 simulations under the null hypothesis (H0). The model is fitted using the Gaussian kernel ψρ(u) = exp(−ρ 2u 2 ). 13 [PITH_FULL_IMAGE:figures/full_fig_p013_2.png] view at source ↗
Figure 3
Figure 3. Quantile–quantile plots of p-values from 5000 simulations under the null hypothesis (H0). The model is fitted using the linear kernel ψρ(u) = max{1 − ρu, 0}. 3.2 Evaluation under the alternative hypothesis Simulations under the alternative hypothesis were conducted to assess the power of the pro￾posed test and its sensitivity to misspecification of the parameter ρ. Two distinct scenarios were considered. Scenario 1.… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Empirical power curves as a function of γ and ρ, with N = 400 based on 1,000 simulations. Data were generated under the alternative hypothesis with ρ = 0.1, while the values of ρ used for model fitting were misspecified. Scenario 2. Data were generated with a higher in…
Figure 5
Figure 5. Figure 5: Empirical power curves as a function of γ and ρ, with N = 400 based on 1,000 simulations. Data were generated under the alternative hypothesis with ρ = 8, while the values of ρ used for model fitting were misspecified. 4 Methylation, genotype and non genetic covariate …
Figure 6
Figure 6. Figure 6: ROC curves showing improved classification performance for the functional interaction mod￾els fitted with different values of ρ (ρ = 0.1, 1, and 10) compared to the reduced logistic model without SNP–methylation interaction (black curve). 18 [PITH_FULL_IMAGE:figures/f…
Figure 7
Figure 7. Figure 7: Empirical power curves as a function of γ1, γ2, and γ3 corresponding respectively to the three scenarios, with sample size N = 355 and 1,000 simulations performed. The black curves correspond to the existing method (LogittestSNPCpG), while the red curves correspond to …

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

88 extracted references

  1. [1]

    Computational statistics & data analysis , volume=

    Interaction models for functional regression , author=. Computational statistics & data analysis , volume=. 2016 , publisher=

    2016

  2. [2]

    Genome biology , volume=

    BSmooth: from whole genome bisulfite sequencing reads to differentially methylated regions , author=. Genome biology , volume=. 2012 , publisher=

    2012

  3. [3]

    Epigenetics and Complex Traits , pages=

    Establishment of tissue-specific epigenetic states during development , author=. Epigenetics and Complex Traits , pages=. 2013 , publisher=

    2013

  4. [4]

    Biometrics , year=

    A novel statistical method for modeling covariate effects in bisulfite sequencing derived measures of DNA methylation , author=. Biometrics , year=

  5. [5]

    Statistical applications in genetics and molecular biology , volume=

    A smoothed EM-algorithm for DNA methylation profiles from sequencing-based methods in cell lines or for a single cell type , author=. Statistical applications in genetics and molecular biology , volume=. 2017 , publisher=

    2017

  6. [6]

    2017 , school=

    A functional data analytic approach for region level differential DNA methylation detection , author=. 2017 , school=

    2017

  7. [7]

    Journal of computational and graphical statistics , volume=

    Penalized functional regression , author=. Journal of computational and graphical statistics , volume=. 2011 , publisher=

    2011

  8. [8]

    2006 , publisher=

    Generalized additive models: an introduction with R , author=. 2006 , publisher=

    2006

Show all 88 references
  1. [9]

    2009 , publisher=

    Functional Data Analysis with R and MATLAB , author=. 2009 , publisher=

    2009

  2. [10]

    Computational Statistics , volume=

    Penalized function-on-function regression , author=. Computational Statistics , volume=. 2015 , publisher=

    2015

  3. [11]

    Biometrika , volume=

    On p-values for smooth components of an extended generalized additive model , author=. Biometrika , volume=. 2013 , publisher=

    2013

  4. [12]

    Statistica Sinica , volume=

    A note on a nonparametric regression test through penalized splines , author=. Statistica Sinica , volume=. 2014 , publisher=

    2014

  5. [13]

    2005 , publisher =

    Functional Data Analysis , author =. 2005 , publisher =

    2005

  6. [14]

    Journal of the Royal Statistical Society: Series B (Statistical Methodology) , volume=

    Smoothing parameter selection for a class of semiparametric linear models , author=. Journal of the Royal Statistical Society: Series B (Statistical Methodology) , volume=. 2009 , publisher=

    2009

  7. [15]

    Nat Genet , volume=

    Generalized loss of stability of epigenetic domains across cancer types , author=. Nat Genet , volume=

  8. [16]

    Diabetes , volume=

    Altered DNA methylation and differential expression of genes influencing metabolism and inflammation in adipose tissue from subjects with type 2 diabetes , author=. Diabetes , volume=. 2014 , publisher=

    2014

  9. [17]

    Nature neuroscience , volume=

    Alzheimer's disease: early alterations in brain DNA methylation at ANK1, BIN1, RHBDF2 and other loci , author=. Nature neuroscience , volume=. 2014 , publisher=

    2014

  10. [18]

    Nature biotechnology , volume=

    Epigenome-wide association data implicate DNA methylation as an intermediary of genetic risk in rheumatoid arthritis , author=. Nature biotechnology , volume=. 2013 , publisher=

    2013

  11. [19]

    Genome biology , volume=

    DNA methylation age of human tissues and cell types , author=. Genome biology , volume=. 2013 , publisher=

    2013

  12. [20]

    JAMA oncology , volume=

    Correlation of smoking-associated DNA methylation changes in buccal cells with DNA methylation changes in epithelial cancer , author=. JAMA oncology , volume=. 2015 , publisher=

    2015

  13. [21]

    , author=

    A genomic sequencing protocol that yields a positive display of 5-methylcytosine residues in individual DNA strands. , author=. Proceedings of the National Academy of Sciences , volume=. 1992 , publisher=

    1992

  14. [22]

    Nucleic acids research , volume=

    Post-conversion targeted capture of modified cytosines in mammalian and plant genomes , author=. Nucleic acids research , volume=. 2015 , publisher=

    2015

  15. [23]

    Nature communications , volume=

    Characterization of functional methylomes by next-generation capture sequencing identifies novel disease-associated variants , author=. Nature communications , volume=. 2015 , publisher=

    2015

  16. [24]

    Journal of the Royal Statistical Society: Series B (Statistical Methodology) , volume=

    On variable bandwidth selection in local polynomial regression , author=. Journal of the Royal Statistical Society: Series B (Statistical Methodology) , volume=. 2000 , publisher=

    2000

  17. [25]

    BMC proceedings , volume=

    Adaptive bandwidth kernel density estimation for next-generation sequencing data , author=. BMC proceedings , volume=. 2013 , organization=

    2013

  18. [26]

    science , volume=

    The sequence of the human genome , author=. science , volume=. 2001 , publisher=

    2001

  19. [27]

    Nucleic acids research , volume=

    CRISPR/Cas9 systems have off-target activity with insertions or deletions between target DNA and guide RNA sequences , author=. Nucleic acids research , volume=. 2014 , publisher=

    2014

  20. [28]

    Nature Reviews Genetics , volume=

    Structural variation in the human genome , author=. Nature Reviews Genetics , volume=. 2006 , publisher=

    2006

  21. [29]

    Nature reviews genetics , volume=

    Environmental epigenomics and disease susceptibility , author=. Nature reviews genetics , volume=. 2007 , publisher=

    2007

  22. [30]

    BMC genomics , volume=

    Genetic analysis of DNA methylation and gene expression levels in whole blood of healthy human subjects , author=. BMC genomics , volume=. 2012 , publisher=

    2012

  23. [31]

    Human molecular genetics , volume=

    Genome-wide association studies in cancer , author=. Human molecular genetics , volume=. 2008 , publisher=

    2008

  24. [32]

    Clinical chemistry , volume=

    Genomewide association studies in cardiovascular disease—an update 2011 , author=. Clinical chemistry , volume=. 2012 , publisher=

    2011

  25. [33]

    elife , volume=

    Passive and active DNA methylation and the interplay with genetic variation in gene regulation , author=. elife , volume=. 2013 , publisher=

    2013

  26. [34]

    Frontiers in Cell and Developmental Biology , volume=

    The interaction analysis of SNP variants and DNA methylation identifies novel methylated pathogenesis genes in congenital heart diseases , author=. Frontiers in Cell and Developmental Biology , volume=. 2021 , publisher=

    2021

  27. [35]

    Clinical epigenetics , volume=

    The interaction of genetic variants and DNA methylation of the interleukin-4 receptor gene increase the risk of asthma at age 18 years , author=. Clinical epigenetics , volume=. 2013 , publisher=

    2013

  28. [36]

    Human genetics , volume=

    Gender specific differences in levels of DNA methylation at selected loci from human total blood: a tendency toward higher methylation levels in males , author=. Human genetics , volume=. 2007 , publisher=

    2007

  29. [37]

    2003 , publisher=

    Semiparametric regression , author=. 2003 , publisher=

    2003

  30. [38]

    2017 , publisher=

    Generalized additive models: an introduction with R , author=. 2017 , publisher=

    2017

  31. [39]

    Statistical science , volume=

    Flexible smoothing with B-splines and penalties , author=. Statistical science , volume=. 1996 , publisher=

    1996

  32. [40]

    Genome medicine , volume=

    Many obesity-associated SNPs strongly associate with DNA methylation changes at proximal promoters and enhancers , author=. Genome medicine , volume=. 2015 , publisher=

    2015

  33. [41]

    Cancer epidemiology, biomarkers & prevention , volume=

    Peripheral blood immune cell methylation profiles are associated with nonhematopoietic cancers , author=. Cancer epidemiology, biomarkers & prevention , volume=. 2012 , publisher=

    2012

  34. [42]

    BMC proceedings , volume=

    Epigenome wide association study of SNP--CpG interactions on changes in triglyceride levels after pharmaceutical intervention: a GAW20 analysis , author=. BMC proceedings , volume=. 2018 , organization=

    2018

  35. [43]

    Frontiers in Genetics , volume=

    Genetic regulation of newborn telomere length is mediated and modified by DNA methylation , author=. Frontiers in Genetics , volume=. 2022 , publisher=

    2022

  36. [44]

    The Lancet , volume=

    DNA methylation and body-mass index: a genome-wide analysis , author=. The Lancet , volume=. 2014 , publisher=

    2014

  37. [45]

    Nature Genetics , volume=

    Six new loci associated with body mass index highlight a neuronal influence on body weight regulation , author=. Nature Genetics , volume=. 2009 , publisher=

    2009

  38. [46]

    Journal of the American Statistical Association , volume=

    Smoothing parameter and model selection for general smooth models , author=. Journal of the American Statistical Association , volume=. 2016 , publisher=

    2016

  39. [47]

    Genome biology , volume=

    An evaluation of methods correcting for cell-type heterogeneity in DNA methylation studies , author=. Genome biology , volume=. 2016 , publisher=

    2016

  40. [48]

    Briefings in bioinformatics , volume=

    Methodological aspects of whole-genome bisulfite sequencing analysis , author=. Briefings in bioinformatics , volume=. 2015 , publisher=

    2015

  41. [49]

    Nucleic acids research , volume=

    The NHGRI-EBI GWAS Catalog of published genome-wide association studies, targeted arrays and summary statistics 2019 , author=. Nucleic acids research , volume=. 2019 , publisher=

    2019

  42. [50]

    Nucleic acids research , volume=

    The new NHGRI-EBI Catalog of published genome-wide association studies (GWAS Catalog) , author=. Nucleic acids research , volume=. 2017 , publisher=

    2017

  43. [51]

    2008 , publisher=

    Personal genomes: The case of the missing heritability , author=. 2008 , publisher=

    2008

  44. [52]

    Human genomics , volume=

    Epigenetic inheritance and the missing heritability , author=. Human genomics , volume=. 2015 , publisher=

    2015

  45. [53]

    Nature genetics , volume=

    Identifying loci affecting trait variability and detecting interactions in genome-wide association studies , author=. Nature genetics , volume=. 2018 , publisher=

    2018

  46. [54]

    Neuropsychopharmacology , volume=

    Epigenetic mechanisms for the early environmental regulation of hippocampal glucocorticoid receptor gene expression in rodents and humans , author=. Neuropsychopharmacology , volume=. 2013 , publisher=

    2013

  47. [55]

    Epigenetics and complex traits , pages=

    Impact of the early-life environment on the epigenome and behavioral development , author=. Epigenetics and complex traits , pages=. 2013 , publisher=

    2013

  48. [56]

    Epigenetics , volume=

    SNPs located at CpG sites modulate genome-epigenome interaction , author=. Epigenetics , volume=. 2013 , publisher=

    2013

  49. [57]

    Nature genetics , volume=

    Variation is the spice of life , author=. Nature genetics , volume=. 2001 , publisher=

    2001

  50. [58]

    Nature , volume=

    Shotgun bisulphite sequencing of the Arabidopsis genome reveals DNA methylation patterning , author=. Nature , volume=. 2008 , publisher=

    2008

  51. [59]

    Nature protocols , volume=

    Preparation of reduced representation bisulfite sequencing libraries for genome-scale DNA methylation profiling , author=. Nature protocols , volume=. 2011 , publisher=

    2011

  52. [60]

    Biostatistics , volume=

    Detection and accurate false discovery rate control of differentially methylated regions from whole genome bisulfite sequencing , author=. Biostatistics , volume=. 2019 , publisher=

    2019

  53. [61]

    Bioinformatics , volume=

    Detection of significantly differentially methylated regions in targeted bisulfite sequencing data , author=. Bioinformatics , volume=. 2013 , publisher=

    2013

  54. [62]

    Genomics , volume=

    Nucleotide distance influences co-methylation between nearby CpG sites , author=. Genomics , volume=. 2020 , publisher=

    2020

  55. [63]

    Nature Reviews Genetics , volume=

    Epigenome-wide association studies for common human diseases , author=. Nature Reviews Genetics , volume=. 2011 , publisher=

    2011

  56. [64]

    Nucleic acids research , volume=

    EWAS Atlas: a curated knowledgebase of epigenome-wide association studies , author=. Nucleic acids research , volume=. 2019 , publisher=

    2019

  57. [65]

    BMC genomics , volume=

    Accounting for multiple comparisons in a genome-wide association study (GWAS) , author=. BMC genomics , volume=. 2010 , publisher=

    2010

  58. [66]

    The American journal of human genetics , volume=

    PLINK: a tool set for whole-genome association and population-based linkage analyses , author=. The American journal of human genetics , volume=. 2007 , publisher=

    2007

  59. [67]

    Nature , volume=

    Linkage disequilibrium in the human genome , author=. Nature , volume=. 2001 , publisher=

    2001

  60. [68]

    Genome biology , volume=

    DNA methylation signatures of chronic low-grade inflammation are associated with complex diseases , author=. Genome biology , volume=. 2016 , publisher=

    2016

  61. [69]

    PLoS medicine , volume=

    Epigenetic epidemiology of common complex disease: prospects for prediction, prevention, and treatment , author=. PLoS medicine , volume=. 2010 , publisher=

    2010

  62. [70]

    Nature genetics , volume=

    Mixed linear model approach adapted for genome-wide association studies , author=. Nature genetics , volume=. 2010 , publisher=

    2010

  63. [71]

    Nature methods , volume=

    Improved linear mixed models for genome-wide association studies , author=. Nature methods , volume=. 2012 , publisher=

    2012

  64. [72]

    2017 , publisher=

    Introduction to functional data analysis , author=. 2017 , publisher=

    2017

  65. [73]

    1994 , publisher=

    Kernel smoothing , author=. 1994 , publisher=

    1994

  66. [74]

    Bioinformatics , volume=

    Functional data analysis for computational biology , author=. Bioinformatics , volume=. 2019 , publisher=

    2019

  67. [75]

    R package version , volume=

    Package ‘mgcv’ , author=. R package version , volume=

  68. [76]

    Journal of the Royal Statistical Society Series B: Statistical Methodology , volume=

    Laplace approximation of high dimensional integrals , author=. Journal of the Royal Statistical Society Series B: Statistical Methodology , volume=. 1995 , publisher=

    1995

  69. [77]

    Journal of the American Statistical Association , volume=

    Generalized additive models: some applications , author=. Journal of the American Statistical Association , volume=. 1987 , publisher=

    1987

  70. [78]

    Theory of Probability & Its Applications , volume=

    On estimating regression , author=. Theory of Probability & Its Applications , volume=. 1964 , publisher=

    1964

  71. [79]

    1978 , publisher=

    A practical guide to splines , author=. 1978 , publisher=

    1978

  72. [80]

    Nature genetics , volume=

    DNA methylation profiling of human chromosomes 6, 20 and 22 , author=. Nature genetics , volume=. 2006 , publisher=

    2006

  73. [81]

    Numerische mathematik , volume=

    Smoothing by spline functions , author=. Numerische mathematik , volume=. 1967 , publisher=

    1967

  74. [82]

    2024 , publisher=

    Functional data analysis with R , author=. 2024 , publisher=

    2024

  75. [83]

    Statistical science , pages=

    A statistical perspective on ill-posed inverse problems , author=. Statistical science , pages=. 1986 , publisher=

    1986

  76. [84]

    1990 , publisher=

    Spline models for observational data , author=. 1990 , publisher=

    1990

  77. [85]

    Biometrika , volume=

    On the existence of maximum likelihood estimates in logistic regression models , author=. Biometrika , volume=. 1984 , publisher=

    1984

  78. [86]

    2019 , publisher=

    Generalized linear models , author=. 2019 , publisher=

    2019

  79. [87]

    1989 , edition=

    Generalized Linear Models , author=. 1989 , edition=

    1989

  80. [88]

    Statistics in Medicine , volume=

    A Functional Approach to Testing Overall Effect of Interaction Between DNA Methylation and SNPs , author=. Statistics in Medicine , volume=. 2026 , publisher=

    2026

Pith tools

Reviewed July 13, 2026 · model on record in the stance chip above.