Adiv Johnson at ARDD2026: Buccal methylomic data to generate novel insights into epigenetic aging

Adiv Johnson at ARDD2026: Buccal methylomic data to generate novel insights into epigenetic aging

by Adiv Johnson · View the channel on PSI

Life & Natural Systems Biology PSBiology, life sciencesPSDMolecular biology
🎙 Adiv Johnson 👥 9K 📅 October 16, 2026 ⏱ 19 min 👁 7 📄 original study 🧭 2026-10-16
Available in: English (current) Français

Keywords

epigenetic clockbuccal swabmethylation sequencingbiomarker proxiestechnical noise

Summary

This talk from ARDD2026 presents recent work by Tally Health on using buccal (cheek swab) DNA methylation data to advance epigenetic aging research. The speaker outlines the development of CheekAge, a next-generation clock built on 8,045 beta testers with a large dataset of more than 25,000 DNA methylation samples, including nearly 10,000 longitudinal and over 600 replicates. A key focus is on reducing technical noise: a comparison between Illumina methylation arrays and Twist methylation sequencing shows that arrays are more reproducible at the CpG level, while principal component clocks suppress noise effectively on both platforms. The talk also addresses explainability, arguing that epigenetic biomarker proxies (predicting proteins, metabolites, clinical variables from methylation) are more actionable than gene-annotated pathway approaches. A proof-of-concept deep neural network trained on 10,000 principal components predicts lifestyle and health factors from survey data, achieving an ICC of 0.996 and a mean absolute error of ~2 years. Associations with BMI, omega-3 intake, ultra-processed food, and other factors are reported. The work demonstrates that large datasets with rich metadata enable clocks that combine accuracy, explainability, and reproducibility.

179 words

Critical Evaluation

Value of the Information & Strength of the Argument

The value lies in the systematic comparison of methylation platforms and the emphasis on noise suppression, which is critical for clinical utility. The argumentation is solid: the speaker uses large sample sizes, technical replicates, and rigorous statistical metrics (ICC, R², mean absolute error). The introduction of epigenetic biomarker proxies as standalone biomarkers is a strong conceptual contribution, supported by a semaglutide study showing modifiable predictions. The proof-of-concept model is well-validated with cross-validation and longitudinal agreement, and the logic from hypothesis to test to conclusion is clear. However, the presentation is a conference talk, so some details of methods and limitations are condensed; still, the overall claims are backed by published and preprint papers.

Scientific Rigor, Source Quality, Title Accuracy

The speaker cites peer-reviewed publications (e.g., in GeroScience) and mentions public data deposition with a DOI. The sources are primarily from the speaker’s own work and collaborators (Tally Health, TruDiagnostic, UCSD), which is appropriate for a presentation of original research. The title is accurate and not overstated. The talk includes commentary from a question from Steve (likely Steve Horvath) about smoking association, showing transparency about unexpected findings. The description lists conference sponsors, but these are not used as scientific sources. Overall, the rigor is high, typical for a specialized academic audience, and the title-content alignment is strong.

225 words

Title / Content Match

The title accurately describes the talk's focus on using buccal (cheek swab) methylomic data for epigenetic aging insights.

Quality & Reliability

8/10

The presentation is based on peer-reviewed research and large datasets (25,000+ samples), with public data availability. The speaker discloses methods and limitations, and the claims are supported by multiple correlations and validation steps.

Key Moments

Markers derived by PSI from the transcript: the creator did not define chapters.

Cited Sources

  • AgingPHarma.org — Website for the Aging Research and Drug Discovery (ARDD) meeting, hosted by Insilico Medicine.
  • ARDD on LinkedIn — Official LinkedIn page for the ARDD conference.

Contribution & Novelties

The talk provides a clear demonstration that Illumina methylation arrays are less noisy than Twist methylation sequencing at the CpG level, while principal component-based clocks can mitigate noise on both platforms. It also introduces epigenetic biomarker proxies as valuable standalone aging biomarkers and presents a proof-of-concept model that combines accuracy, explainability, and reproducibility using buccal samples. The approach of training a deep network on survey data and principal components is novel in the field.

Pour aller plus loin :

  • Epigenetic clock — Core concept behind the talk; various clocks and their limitations.
  • DNA methylation — The molecular basis of the data used; key to understanding methylomic profiling.
  • Principal component analysis — The dimensionality reduction technique used to suppress noise and build the new clock.

124 words

Radar Profile

The profile is strong on technical level and information quantity, reflecting the speaker's deep expertise and comprehensive dataset. Quality and reliability are also high, though the presentation format limits depth of methods. Overall, a balanced, technically advanced talk.

Reliability 7/10