Deep Learning Model Melody Reads DNA Code to Predict Methylation Patterns

The work was published in Nature Communications in 2026 by Jin and colleagues. It is a computational modelling study, not a trial in people. The researchers built a deep learning model called Melody and trained it on methylation maps for 39 normal human cell types. The study does not test any consumer age test.

Methylation is a chemical mark added to DNA. It helps control which genes are switched on. According to the paper, methylation patterns also reflect biological age and exposure history, and they are used in cancer diagnostics, disease screening and age estimation. What has been less clear is how far the DNA sequence itself decides where methylation appears at a given spot.

Melody reads a stretch of 10,000 DNA letters and predicts methylation levels at individual sites within it. The authors say earlier models used much shorter windows, such as 41 letters. They argue that shorter windows miss distant signals, like nearby CpG islands, that influence local methylation. In their tests, around 10,000 letters worked best. A 2,000-letter window did worse, and windows of 20,000 to 50,000 letters added redundant sequence.

The results were stronger than those of the competing methods the team tested, which included INTERACT, CpGenie and iDNA-ABT. Across 39 datasets, the multitissue version, Melody-MT, reached a mean Spearman correlation of 0.723 on one test set and 0.645 on a held-out chromosome. The strongest competing baselines scored 0.590 and 0.584. Spearman correlation measures how well predicted and measured values rank in the same order, where 1 is a perfect match. Across the two Melody variants, the authors report an average gain of about 17% in that measure.

The model also held up in regions with low methylation, where the authors say most baselines got worse. On individual chromosomes, Melody showed about a 7% relative improvement in Pearson correlation over the best competing model, according to the paper.

The team also tested whether Melody could predict the effect of genetic variants that shift methylation, known as meQTLs. The study says it generalizes to this task and identifies sequence motifs linked to methylation variability. The authors note that no standard benchmark existed for this, so they built their own meQTL benchmark datasets. The excerpt available to us does not give the meQTL results in detail.

A companion version, Melody-G, adds data from single-cell RNA sequencing, which measures which genes a cell is using. This lets it estimate methylation in cell types it was not trained on. The study describes this as a way to infer methylation states in previously unseen cell types. The excerpt we reviewed does not report how accurate those estimates were.

What does this mean for people curious about age tests? Mainly, it is basic science. It suggests that the DNA sequence carries more information about methylation than shorter-window models captured. That could help researchers interpret why methylation differs between tissues and why some genetic variants change it. The study does not show that Melody improves age estimates or health predictions, and it does not show that any lifestyle change alters methylation. All the models were tested on data from normal human cell types, not on disease or ageing outcomes.

The authors also released a web server for viewing predicted methylation across tissues. The press source page we were given could not be read, so this article relies on the study itself.

What to watch next

Whether independent groups confirm Melody's results on new tissues, and whether it is tested on disease samples or age-related methylation change.

Removing Brain Immune Cells Restored Two Hours of Sleep in Alzheimer's Mice Gut Bacteria and Chemicals Differ in Mild Cognitive Impairment, Small Study Finds Starting Statins Soon After Diabetes Diagnosis Linked to Less Dementia Flavourings, Colourings and Sweeteners Linked to Higher Depression Risk Older Adults on Self-Chosen Vegan Diet Lost Leg Muscle in 12 Weeks Topical Vitamin K2 Cream Improved Wrinkles and Firmness in Small Trial Staying Active Over Years Linked to 44% Lower Frailty Risk East Germans Who Faced Covert State Repression Still Report More Stress Four Cardio Styles Compared: High-Intensity Intervals Beat Moderate Intervals on Fitness AI Predicted Brain Amyloid From Blood Tests, but Faltered on New Patients Boosting Gene ZNF865 With CRISPR Reversed Aging Signs in Degenerated Spinal Disc Cells Review of 26 Cohorts Links Time-Restricted Eating to Better Blood Sugar Years of Fitbit Data Link Steady Daily Rhythms to Slower Biological Aging Review of Nine Meta-Analyses Links Hearing Loss to 43% Higher Dementia Odds UK Biobank Study of 170,000 Women Links HRT to Higher Cardiovascular Risk Small Trial Finds Ultra-Processing Alone Did Not Change Short-Term Satiety Combining Cardio and Strength Training Beats Cardio Alone in Heart Rehab 20-Gram Protein Dose Failed to Boost Muscle Building in Older Men Mediterranean Diet Linked to Slightly Younger DNA-Based Aging Readings in Study Gut Microbes May Shape How Environmental Toxicants Affect the Brain, Review Says Review Says Simply Raising NAD+ or SIRT1 May Not Slow Ageing Review Calls Chronic Stress Before Surgery a Possible Modifiable Heart Risk Review Maps How CRISPR Delivery Could Target Aging's Hallmarks Deep Learning Model Melody Reads DNA Code to Predict Methylation Patterns Mostly-Easy Polarized Training Beat Threshold Workouts on Endurance Markers, Not VO2max Trump Executive Order Creates New NIH Division and Task Force on Women's Health Review of 46 Studies Finds Ultra-Processed Food Heart Research Methodologically Weak Review Finds No Consistent Epigenetic Age Reversal From Antioxidant Supplements Time-Restricted Eating Raises Ketones, Modestly Calms Inflammation, Review Finds Gut-Made Short-Chain Fatty Acids Linked to Brain Protection in Low Blood Flow