pith. machine review for the scientific record. sign in

arxiv: 2605.10541 · v1 · submitted 2026-05-11 · 💻 cs.AI · cs.LG

Recognition: 2 theorem links

· Lean Theorem

Bridging Sequence and Graph Structure for Epigenetic Age Prediction

Yao Li , Xikun Zhang , Xiaotao Shen , Sonika Tyagi , Xin Zheng , Jiaxing Huang , Feng Xia
Authors on Pith no claims yet

Pith reviewed 2026-05-12 04:21 UTC · model grok-4.3

classification 💻 cs.AI cs.LG
keywords epigenetic age predictionDNA methylationgraph neural networkssequence featuresgated modulationbiological ageCpG sitesaging research
0
0 comments X

The pith

A gated modulation mechanism integrates eight DNA sequence statistical features with graph convolution to improve epigenetic age prediction.

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

The paper establishes a unified framework that jointly models co-methylation graph structure and site-specific DNA sequence context for epigenetic age prediction. It introduces a lightweight gated mechanism that scales each site's methylation signal according to eight sequence-derived statistical features before graph convolution. Evaluated on 3707 blood samples, this yields a mean absolute error of 3.149 years. The approach demonstrates that handcrafted biological sequence features outperform CNN-based encodings in this setting. More accurate clocks support better tracking of biological aging processes and age-related disease mechanisms.

Core claim

The central claim is that integrating eight-dimensional DNA sequence statistical features through a lightweight gated modulation mechanism to adaptively scale methylation signals according to sequence-determined biological relevance prior to graph convolution produces more accurate epigenetic age estimates than existing graph-only or sequence-only methods.

What carries the argument

The lightweight gated modulation mechanism that uses eight sequence statistical features to adaptively scale each site's methylation signal based on its sequence-determined biological relevance.

If this is right

  • The method reaches a test MAE of 3.149 years on 3707 blood methylation samples.
  • This represents a 12.8% improvement over the strongest graph-based baseline.
  • Biologically informed statistical sequence features outperform CNN-based sequence encoding.
  • Post-hoc interpretability shows CpG density and local adenine frequency exhibit age-dependent importance shifts consistent with known hypermethylation mechanisms.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The success of simple statistical features over learned encodings suggests that incorporating domain knowledge can be more data-efficient than end-to-end deep learning for genomic graphs.
  • The modulation approach could be adapted to predict other traits from methylation data by swapping in different sequence-derived priors.
  • The identified age-dependent feature shifts point to testable hypotheses about how promoter CpG density influences methylation drift over time.

Load-bearing premise

The gated modulation driven by the eight sequence features genuinely captures biological relevance in a way that generalizes rather than overfits the 3707-sample dataset.

What would settle it

Failure to achieve lower MAE than strong graph baselines on a new independent cohort of blood methylation samples would show the claimed improvement does not hold.

Figures

Figures reproduced from arXiv: 2605.10541 by Feng Xia, Jiaxing Huang, Sonika Tyagi, Xiaotao Shen, Xikun Zhang, Xin Zheng, Yao Li.

Figure 1
Figure 1. Figure 1: Overview of the proposed sequence–graph integration framework. Eight-dimensional [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Predicted vs. true age for the three variants ( [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Test MAE by age group (n = 756). Age group performance [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Temporal analysis of sequence feature importance scores. [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Age distribution of the blood methylation dataset ( [PITH_FULL_IMAGE:figures/full_fig_p013_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Linear regression fits of sequence feature importance scores across chronological age. [PITH_FULL_IMAGE:figures/full_fig_p014_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Top 10 CpG sites with increasing (a) and decreasing (b) node importance scores across [PITH_FULL_IMAGE:figures/full_fig_p015_7.png] view at source ↗
read the original abstract

Epigenetic clocks based on DNA methylation have emerged as powerful tools for estimating biological age, with broad applications in aging research, age-related disease studies, and longevity science. Despite advances across machine learning approaches to epigenetic age prediction, spanning penalised linear regression, deep feedforward networks, residual architectures, and graph neural networks, no existing method jointly models co-methylation graph structure and site-specific DNA sequence context within a unified framework. We propose a unified sequence--graph integration framework for epigenetic age prediction that addresses this gap, integrating eight-dimensional DNA sequence statistical features through a lightweight gated modulation mechanism that adaptively scales each site's methylation signal according to its sequence-determined biological relevance prior to graph convolution. Evaluated on 3,707 blood methylation samples against a comprehensive set of baselines, our method achieves a test MAE of 3.149 years, a 12.8\% improvement over the strongest graph-based baseline. Biologically informed statistical features outperform CNN-based sequence encoding, demonstrating that handcrafted sequence features are more effective than end-to-end learned representations in this data regime. Post-hoc interpretability analysis identifies CpG density and local adenine frequency as features with age-dependent importance shifts, consistent with known mechanisms of age-related hypermethylation at CpG-dense promoter regions. Our code is at https://github.com/yaoli2022/graphage-seq.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript proposes a unified sequence-graph framework for epigenetic age prediction from DNA methylation data. It extracts eight-dimensional statistical features from DNA sequences around CpG sites and uses a lightweight gated modulation mechanism to adaptively scale methylation signals based on sequence context before applying graph convolutions on co-methylation graphs. On a dataset of 3,707 blood samples, the method reports a test MAE of 3.149 years (12.8% improvement over the strongest graph baseline), claims superiority of handcrafted features over CNN sequence encoders, and provides post-hoc interpretability linking CpG density and adenine frequency to age-related changes. Code is made available.

Significance. If the performance gains prove robust, the work could advance epigenetic clock modeling by showing benefits of explicit sequence-graph integration over pure graph or sequence approaches. The public code release supports reproducibility, a clear strength. However, the empirical claims rest on a single internal test set without external cohorts, so broader significance depends on verification of generalization.

major comments (2)
  1. [Abstract] Abstract: The central claim of a test MAE of 3.149 years and 12.8% improvement over the strongest graph baseline is stated without any mention of train/test split details, baseline implementation specifics, number of random seeds or runs, error bars, or statistical significance tests. These omissions make it impossible to assess whether the reported gain reflects genuine sequence-graph integration or a favorable split on the 3,707-sample cohort.
  2. [Abstract] Abstract and method description: The lightweight gated modulation mechanism is presented as adaptively scaling methylation signals according to sequence-determined biological relevance, yet no ablation isolating the modulation weights from the eight handcrafted features is described, nor is there analysis showing that the modulation produces better generalization than simply using the features as additional node attributes. With high-dimensional CpG graphs and limited samples, this leaves open whether the mechanism drives the gain or amplifies dataset-specific correlations.
minor comments (2)
  1. [Abstract] The abstract states that 'biologically informed statistical features outperform CNN-based sequence encoding' but does not specify the CNN architecture, training regime, or whether the comparison used identical graph backbones; this should be clarified for direct comparability.
  2. [Abstract] The GitHub link is provided, but the manuscript should explicitly state which commit or release tag corresponds to the exact code and hyperparameters that produced the reported MAE numbers.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments on our manuscript. We address each major comment point by point below, providing clarifications based on the full paper and indicating where revisions will be made to improve clarity and rigor.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The central claim of a test MAE of 3.149 years and 12.8% improvement over the strongest graph baseline is stated without any mention of train/test split details, baseline implementation specifics, number of random seeds or runs, error bars, or statistical significance tests. These omissions make it impossible to assess whether the reported gain reflects genuine sequence-graph integration or a favorable split on the 3,707-sample cohort.

    Authors: We acknowledge that the abstract's brevity omits these details. The full manuscript's Experiments section specifies the train/test split procedure on the 3,707-sample cohort, the reimplementation of baselines following their original descriptions, the use of multiple random seeds with reported means and standard deviations (error bars), and the application of statistical significance tests. To address the referee's concern directly, we will revise the abstract to include a concise reference to the evaluation protocol. revision: yes

  2. Referee: [Abstract] Abstract and method description: The lightweight gated modulation mechanism is presented as adaptively scaling methylation signals according to sequence-determined biological relevance, yet no ablation isolating the modulation weights from the eight handcrafted features is described, nor is there analysis showing that the modulation produces better generalization than simply using the features as additional node attributes. With high-dimensional CpG graphs and limited samples, this leaves open whether the mechanism drives the gain or amplifies dataset-specific correlations.

    Authors: This observation is correct; the manuscript does not include a dedicated ablation that isolates the gated modulation by comparing it to a non-gated variant in which the eight features are directly concatenated as additional node attributes. While the paper demonstrates the overall benefits of handcrafted features and sequence-graph integration through comparisons to CNN encoders and graph baselines, an explicit ablation of this form is absent. We will add this ablation study to the revised manuscript, reporting test-set performance for both the full gated model and the direct-concatenation variant to quantify the contribution of adaptive scaling. revision: yes

Circularity Check

0 steps flagged

No circularity in empirical performance claims or derivation

full rationale

The paper's central claims consist of an empirical test MAE of 3.149 years and a 12.8% improvement over graph baselines, measured on held-out data from 3707 samples. The proposed integration of eight-dimensional sequence features via gated modulation prior to graph convolution is a modeling choice whose effectiveness is evaluated externally rather than defined into the result. No equations reduce the reported metrics to fitted parameters by construction, no self-citations serve as load-bearing uniqueness theorems, and no ansatz or renaming is presented as a derivation. The post-hoc interpretability analysis is consistent with external biology but does not substitute for the performance numbers. This is a standard empirical ML evaluation with no detectable circular steps.

Axiom & Free-Parameter Ledger

2 free parameters · 1 axioms · 1 invented entities

The framework rests on the choice of eight handcrafted sequence features and the design of the gated modulation; these are domain-motivated but not derived from first principles within the paper.

free parameters (2)
  • gated modulation weights
    Learnable parameters in the lightweight gate that scale methylation signals; fitted during training on the methylation samples.
  • graph convolution hyperparameters
    Standard GNN training choices such as layer count and aggregation functions that affect the final MAE.
axioms (1)
  • domain assumption Eight-dimensional statistical features extracted from local DNA sequence provide a valid biological relevance prior for modulating methylation signals
    Invoked when the gated mechanism is introduced; no derivation shows why these particular statistics are optimal.
invented entities (1)
  • lightweight gated modulation mechanism no independent evidence
    purpose: Adaptively scale each site's methylation signal according to sequence features before graph convolution
    New architectural component introduced to bridge sequence and graph; no independent evidence outside the model performance is supplied.

pith-pipeline@v0.9.0 · 5554 in / 1467 out tokens · 68913 ms · 2026-05-12T04:21:42.794689+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

39 extracted references · 39 canonical work pages

  1. [1]

    Epigenetic regulation of gene expression: how the genome integrates intrinsic and environmental signals.Nature Genetics, 33(Suppl):245–254, 2003

    Rudolf Jaenisch and Adrian Bird. Epigenetic regulation of gene expression: how the genome integrates intrinsic and environmental signals.Nature Genetics, 33(Suppl):245–254, 2003. doi: 10.1038/ng1089

    work page doi:10.1038/ng1089 2003

  2. [2]

    DNA methylation age of human tissues and cell types.Genome Biology, 14(10): R115, 2013

    Steve Horvath. DNA methylation age of human tissues and cell types.Genome Biology, 14(10): R115, 2013. doi: 10.1186/gb-2013-14-10-r115

    work page doi:10.1186/gb-2013-14-10-r115 2013

  3. [3]

    Genome-wide methylation profiles reveal quantitative views of human aging rates.Molecular Cell, 49(2):359–367, 2013

    Gregory Hannum, Justin Guinney, Ling Zhao, Li Zhang, Guy Hughes, SriniVas Sadda, Brandy Klotzle, Marina Bibikova, Jian-Bing Fan, Yuan Gao, Rob Deconde, Menzies Chen, Indika Rajapakse, Stephen Friend, Trey Ideker, and Kang Zhang. Genome-wide methylation profiles reveal quantitative views of human aging rates.Molecular Cell, 49(2):359–367, 2013. doi: 10.101...

    work page doi:10.1016/j.molcel.2012.10.016 2013

  4. [4]

    Blasco, Linda Partridge, Manuel Serrano, and Guido Kroemer

    Carlos López-Otín, Maria A. Blasco, Linda Partridge, Manuel Serrano, and Guido Kroemer. The hallmarks of aging.Cell, 153(6):1194–1217, 2013. doi: 10.1016/j.cell.2013.05.039

    work page doi:10.1016/j.cell.2013.05.039 2013

  5. [5]

    Cell172(5), 1091–110717 (2018) https://doi.org/10.1016/j.cell

    Carlos López-Otín, Maria A. Blasco, Linda Partridge, Manuel Serrano, and Guido Kroemer. Hallmarks of aging: An expanding universe.Cell, 186(2):243–278, 2023. doi: 10.1016/j.cell. 2022.11.001

    work page doi:10.1016/j.cell 2023

  6. [6]

    DNA methylation-based biomarkers and the epigenetic clock theory of ageing.Nature Reviews Genetics, 19(6):371–384, 2018

    Steve Horvath and Kenneth Raj. DNA methylation-based biomarkers and the epigenetic clock theory of ageing.Nature Reviews Genetics, 19(6):371–384, 2018. doi: 10.1038/ s41576-018-0004-3

    work page 2018

  7. [7]

    DeepMAge: A methylation aging clock developed with deep learning.Aging and Disease, 12 (5):1252–1262, 2021

    Fedor Galkin, Polina Mamoshina, Kirill Kochetov, Denis Sidorenko, and Alex Zhavoronkov. DeepMAge: A methylation aging clock developed with deep learning.Aging and Disease, 12 (5):1252–1262, 2021. doi: 10.14336/AD.2020.1202

    work page doi:10.14336/ad.2020.1202 2021

  8. [8]

    Lapierre, and Ritambhara Singh

    Lucas Paulo de Lima Camillo, Louis R. Lapierre, and Ritambhara Singh. A pan-tissue DNA- methylation epigenetic clock based on deep learning.npj Aging, 8(1):4, 2022. doi: 10.1038/ s41514-022-00085-y

    work page 2022

  9. [9]

    Sohel Rahman

    Saleh Sakib Ahmed, Nahian Shabab, Md Abul Hassan Samee, and M. Sohel Rahman. GraphAge: Unleashing the power of graph neural network to decode epigenetic aging.PNAS Nexus, 4(6): pgaf177, 2025. doi: 10.1093/pnasnexus/pgaf177

    work page doi:10.1093/pnasnexus/pgaf177 2025

  10. [10]

    Mul- timodal learning with graphs.Nature Machine Intelligence, 5(4):340–350, 2023

    Yasha Ektefaie, George Dasoulas, Ayush Noori, Maha Farhat, and Marinka Zitnik. Mul- timodal learning with graphs.Nature Machine Intelligence, 5(4):340–350, 2023. doi: 10.1038/s42256-023-00624-6

    work page doi:10.1038/s42256-023-00624-6 2023

  11. [11]

    m6Acomet: large-scale functional prediction of individual m6A RNA methylation sites from an RNA co-methylation network.BMC Bioinformatics, 20(1):223, 2019

    Xiangyu Wu, Zhen Wei, Kunqi Chen, Qing Zhang, Jionglong Su, Hui Liu, Lin Zhang, and Jia Meng. m6Acomet: large-scale functional prediction of individual m6A RNA methylation sites from an RNA co-methylation network.BMC Bioinformatics, 20(1):223, 2019. doi: 10.1186/s12859-019-2840-3. 4https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GPL8490 10

    work page doi:10.1186/s12859-019-2840-3 2019

  12. [12]

    The impact of flanking sequence features on DNA CpG methylation.Compu- tational Biology and Chemistry, 92:107480, 2021

    Daniele Santoni. The impact of flanking sequence features on DNA CpG methylation.Compu- tational Biology and Chemistry, 92:107480, 2021. doi: 10.1016/j.compbiolchem.2021.107480

    work page doi:10.1016/j.compbiolchem.2021.107480 2021

  13. [13]

    Nucleotide distance influences co-methylation between nearby CpG sites.Genomics, 112(1):144–150,

    Ornella Affinito, Domenico Palumbo, Annalisa Fierro, Mariella Cuomo, Giulia De Riso, Antonella Monticelli, Gennaro Miele, Lorenzo Chiariotti, and Sergio Cocozza. Nucleotide distance influences co-methylation between nearby CpG sites.Genomics, 112(1):144–150,

    work page

  14. [14]

    doi: 10.1016/j.ygeno.2019.05.007

    work page doi:10.1016/j.ygeno.2019.05.007 2019

  15. [15]

    On over-squashing in message passing neural networks: the impact of width, depth, and topology

    Francesco Di Giovanni, Lorenzo Giusti, Federico Barbero, Giulia Luise, Pietro Liò, and Michael Bronstein. On over-squashing in message passing neural networks: the impact of width, depth, and topology. InProceedings of the 40th International Conference on Machine Learning, ICML’23. JMLR.org, 2023

    work page 2023

  16. [16]

    ResnetAge: A resnet-based DNA methylation age prediction method.Bioengineering, 11(1):34, 2024

    Lijuan Shi, Boquan Hai, Zhejun Kuang, Han Wang, and Jian Zhao. ResnetAge: A resnet-based DNA methylation age prediction method.Bioengineering, 11(1):34, 2024. doi: 10.3390/ bioengineering11010034

    work page 2024

  17. [17]

    Biologically informed deep learning for explainable epigenetic clocks.Scientific Reports, 14:1306, 2024

    Aurel Prosz, Orsolya Pipek, Judit Börcsök, Gergely Palla, Zoltan Szallasi, Sandor Spisak, and Istvan Csabai. Biologically informed deep learning for explainable epigenetic clocks.Scientific Reports, 14:1306, 2024. doi: 10.1038/s41598-023-50495-5

    work page doi:10.1038/s41598-023-50495-5 2024

  18. [18]

    Levine, Ake T

    Morgan E. Levine, Ake T. Lu, Austin Quach, Brian H. Chen, Themistocles L. Assimes, Stefania Bandinelli, Lifang Hou, Andrea A. Baccarelli, James D. Stewart, Yun Li, Eric A. Whitsel, James G. Wilson, Alex P. Reiner, Abraham Aviv, Kurt Lohman, Yongmei Liu, Luigi Ferrucci, and Steve Horvath. An epigenetic biomarker of aging for lifespan and healthspan.Aging, ...

    work page doi:10.18632/aging.101414 2018

  19. [19]

    Function and information content of DNA methylation.Nature, 517(7534): 321–326, 2015

    Dirk Schübeler. Function and information content of DNA methylation.Nature, 517(7534): 321–326, 2015. doi: 10.1038/nature14192

    work page doi:10.1038/nature14192 2015

  20. [20]

    Principal neighbourhood aggregation for graph nets

    Gabriele Corso, Luca Cavalleri, Dominique Beaini, Pietro Liò, and Petar Veliˇckovi´c. Principal neighbourhood aggregation for graph nets. InAdvances in Neural Information Processing Systems, volume 33, pages 13260–13271, 2020

    work page 2020

  21. [21]

    Lee, Wolf Reik, and Oliver Stegle

    Christof Angermueller, Heather J. Lee, Wolf Reik, and Oliver Stegle. DeepCpG: accurate prediction of single-cell DNA methylation states using deep learning.Genome Biology, 18:67,

    work page

  22. [22]

    doi: 10.1186/s13059-017-1189-z

    work page doi:10.1186/s13059-017-1189-z

  23. [23]

    Thomas, Callum Birch Sykes, Michael Wornow, Aman Patel, Clayton Rabideau, Stefano Massaroli, Yoshua Bengio, Stefano Ermon, Stephen A

    Eric Nguyen, Michael Poli, Marjan Faizi, Armin W. Thomas, Callum Birch Sykes, Michael Wornow, Aman Patel, Clayton Rabideau, Stefano Massaroli, Yoshua Bengio, Stefano Ermon, Stephen A. Baccus, and Christopher Ré. HyenaDNA: long-range genomic sequence modeling at single nucleotide resolution. InProceedings of the 37th International Conference on Neural Info...

    work page 2023

  24. [24]

    Davuluri, and Han Liu

    Zhihan Zhou, Yanrong Ji, Weijian Li, Pratik Dutta, Ramana V . Davuluri, and Han Liu. DNABERT-2: Efficient foundation model and benchmark for multi-species genomes. In International Conference on Learning Representations, 2024

    work page 2024

  25. [25]

    Caduceus: bi-directional equivariant long-range DNA sequence modeling

    Yair Schiff, Chia-Hsiang Kao, Aaron Gokaslan, Tri Dao, Albert Gu, and V olodymyr Kuleshov. Caduceus: bi-directional equivariant long-range DNA sequence modeling. InProceedings of the 41st International Conference on Machine Learning, ICML’24. JMLR.org, 2024

    work page 2024

  26. [26]

    BEND: Benchmarking DNA language models on biologically meaningful tasks

    Frederikke Marin, Felix Teufel, Marc Horlacher, Dennis Madsen, Dennis Pultz, Ole Winther, and Wouter Boomsma. BEND: Benchmarking DNA language models on biologically meaningful tasks. InInternational Conference on Learning Representations, pages 15246–15281, 2024

    work page 2024

  27. [27]

    Bozhen Hu, Cheng Tan, Jun Xia, Yue Liu, Lirong Wu, Jiangbin Zheng, Yongjie Xu, Yufei Huang, and Stan Z. Li. Learning complete protein representation by dynamically coupling of sequence and structure. InAdvances in Neural Information Processing Systems, volume 37, pages 137673–137697. Curran Associates, Inc., 2024

    work page 2024

  28. [28]

    NCBI, 2024

    National Center for Biotechnology Information. NCBI, 2024. URL https://www.ncbi.nlm. nih.gov/. 11

    work page 2024

  29. [29]

    EMBL’s European Bioinformatics Institute, 2024

    EMBL-EBI. EMBL’s European Bioinformatics Institute, 2024. URL https://www.ebi.ac. uk/

    work page 2024

  30. [30]

    Self-normalizing neural networks

    Günter Klambauer, Thomas Unterthiner, Andreas Mayr, and Sepp Hochreiter. Self-normalizing neural networks. InProceedings of the 31st International Conference on Neural Information Processing Systems, NIPS’17, pages 972–981, Red Hook, NY , USA, 2017. Curran Associates Inc

    work page 2017

  31. [31]

    Dropout: A simple way to prevent neural networks from overfitting.Journal of Machine Learning Research, 15:1929–1958, 2014

    Nitish Srivastava, Geoffrey Hinton, Alex Krizhevsky, Ilya Sutskever, and Ruslan Salakhutdinov. Dropout: A simple way to prevent neural networks from overfitting.Journal of Machine Learning Research, 15:1929–1958, 2014

    work page 1929

  32. [32]

    Kingma and Jimmy Ba

    Diederik P. Kingma and Jimmy Ba. Adam: A method for stochastic optimization. InInterna- tional Conference on Learning Representations, 2015

    work page 2015

  33. [33]

    Grimm, Wendan Ren, Sabrina Adam, Hide- taka Uryu, Zhi-Min Zhang, Dongliang Chen, Jiekai Yin, Michael Dukatz, Hiwot Anteneh, Renata Z

    Linfeng Gao, Max Emperle, Yiran Guo, Sara A. Grimm, Wendan Ren, Sabrina Adam, Hide- taka Uryu, Zhi-Min Zhang, Dongliang Chen, Jiekai Yin, Michael Dukatz, Hiwot Anteneh, Renata Z. Jurkowska, Jiuwei Lu, Yinsheng Wang, Pavel Bashtrykov, Paul A. Wade, Gang Greg Wang, Albert Jeltsch, and Jikui Song. Comprehensive structure-function characterization of DNMT3B a...

    work page doi:10.1038/s41467-020-17109-4 2020

  34. [34]

    Deaton and Adrian Bird

    Aimée M. Deaton and Adrian Bird. CpG islands and the regulation of transcription.Genes & Development, 25(10):1010–1022, 2011. doi: 10.1101/gad.2037511

    work page doi:10.1101/gad.2037511 2011

  35. [35]

    Aging and epigenetic drift: a vicious cycle.Journal of Clinical Investigation, 124(1):24–29, 2014

    Jean-Pierre Issa. Aging and epigenetic drift: a vicious cycle.Journal of Clinical Investigation, 124(1):24–29, 2014. doi: 10.1172/JCI69735

    work page doi:10.1172/jci69735 2014

  36. [36]

    Sinclair

    Yuancheng Ryan Lu, Xiao Tian, and David A. Sinclair. The information theory of aging.Nature Aging, 3(12):1486–1499, 2023. doi: 10.1038/s43587-023-00527-6

    work page doi:10.1038/s43587-023-00527-6 2023

  37. [37]

    Jones, Sarah J

    Meaghan J. Jones, Sarah J. Goodman, and Michael S. Kobor. DNA methylation and healthy human aging.Aging Cell, 14(6):924–932, 2015. doi: 10.1111/acel.12349

    work page doi:10.1111/acel.12349 2015

  38. [38]

    CpG-island methylation in aging and cancer.Current Topics in Microbiology and Immunology, 249:101–118, 2000

    Jean-Pierre Issa. CpG-island methylation in aging and cancer.Current Topics in Microbiology and Immunology, 249:101–118, 2000. doi: 10.1007/978-3-642-59696-4_7

    work page doi:10.1007/978-3-642-59696-4_7 2000

  39. [39]

    GNNEx- plainer: Generating explanations for graph neural networks

    Zhitao Ying, Dylan Bourgeois, Jiaxuan You, Marinka Zitnik, and Jure Leskovec. GNNEx- plainer: Generating explanations for graph neural networks. InAdvances in Neural Information Processing Systems, volume 32, pages 9240–9251, 2019. A Dataset Details and Preprocessing Sample filtering.Each dataset is filtered to retain only blood tissue samples. Samples wi...

    work page 2019