From correlations to mechanism

Most environmental epigenetics studies (including plenty of ours) are correlational: expose animals to a treatment, profile marks, and look for differences that line up with the environment or with gene expression. That approach has been productive, but it leaves the mechanism as a black box.

Led by Holly Moeller and Roger Nisbet at UC Santa Barbara, this work flips the order. We built an individual-based stochastic simulation in which each organism carries a string of genomic “sites” that are either marked or unmarked, and the environment has its own string describing the best-matched epigenetic state. An organism’s stress is simply the fraction of sites that are mismatched with the environment. The more stressed an organism is, the faster it remodels: marks are added and removed site by site, partly at random and partly “preferentially” at sites the organism can tell are wrong. The model is deliberately agnostic about whether a site is a single CpG, a gene, or a cluster of co-regulated marks, and about whether a mark activates or represses expression.

That framework let us ask three questions, each paired with a look at the empirical evidence in marine invertebrates and a clear statement of what is still unknown.

1. Tracking the environment requires site-level information

Without any information about which sites are mismatched, organisms never do better than chance: stress hovers around 0.5, the expectation for a random epigenome. Give them that information, and populations drop to roughly 15% mismatch within about 100 time steps, spike when the environment shifts, and then recover. The more an organism can separate preferential from random mark removal, the better it does.

This lines up with patterns we and others have seen. Pteropods (Limacina antarctica) lose DNA methylation after acute low pH exposure and return to baseline within 6 days. Heat-stressed Acropora cervicornis methylation profiles diverge and then become indistinguishable from controls 8 days after the stress ends. Oysters, geoduck, and corals all show methylation shifts with acidification. What we still lack is the mechanism: how does an animal translate an environmental cue into a targeted change at the right sites?

2. Adding and removing marks must be coupled

If mark addition and removal get out of balance, an organism either strips its genome bare or marks everything, and either way ends up badly mismatched. Using a simple evolutionary algorithm (the most stressed individual dies, the least stressed reproduces, removal tendencies occasionally mutate), we found that higher addition tendencies always require proportionally higher removal tendencies. The exact balance depends on the environment: settings that call for few marks favor stronger removal.

Biochemically, addition and removal are handled by different enzymes (DNMTs versus TETs for DNA methylation; HATs, HDACs, and histone demethylases for histone marks), and both cost energy. How organisms coordinate these pools is an open question, and one we think ties epigenetics directly to an animal’s energetic state.

3. Epigenetic tracking only works in some environments

We then let the environment shift in predictable ways, varying both how often it changes and how much. Organisms with good environmental information can follow infrequent changes, even big ones (“abrupt jumps”), and frequent changes if they are small (“slow and steady”). Frequent, large changes push stress back toward random.

This leads to a prediction we are excited to test: lineages from highly variable habitats (intertidal zones, estuaries) should lean more heavily on epigenetic regulation than close relatives from stable ones. Oysters are a natural system for this, with Eastern and Pacific oysters living in fluctuating estuaries while relatives such as Crassostrea nippona occupy more stable coastal waters. We also argue that the environmental changes most likely to select for epigenetic responses are reversible ones that recur on the order of a generation or two.

Where the field goes next

The paper closes with a call for experiments designed around mechanism rather than endpoints:

• High-resolution time series after a defined stressor, measuring physiology, gene expression, and epigenetic state together, repeated across lineages and species.
• Quantifying the epigenetic machinery (DNMT, TET, HAT, HDAC, and HDM abundance and activity) to see how addition and removal are coupled.
• Phylogenetically controlled comparisons of lineages from stable versus variable environments, using common garden and reciprocal transplant designs.

For our lab, this is a roadmap. It pushes us toward designs that manipulate the timing and predictability of environmental change, measure mark removal as carefully as mark addition, and follow individuals through time rather than relying on single snapshots.

A collaborative effort

This paper came out of the NSF Understanding the Rules of Life: Epigenetics program, bringing together modelers and empiricists: Holly Moeller and Roger Nisbet (UC Santa Barbara), Hollie Putnam (University of Rhode Island), Ross Cunning (Shedd Aquarium), Jose Eirin-Lopez (Florida International University), and our group at the UW School of Aquatic and Fishery Sciences. Bringing explicit mathematical frameworks together with empirical data is, we argue, exactly what the field needs to move forward.

Learn more

• Full paper (open access): https://doi.org/10.1093/eep/dvag032
• Model code: https://doi.org/10.5281/zenodo.23108257
• Summary video and more: Genefish post
• Preprint: https://doi.org/10.64898/2026.01.08.698487

Figures reproduced from Moeller et al. (2026) under a CC BY-NC 4.0 license.

Hannah Nowers — Cracking the case of the wasting sea stars 🌟

Sea star wasting disease has wiped out 20+ species from Alaska to Baja, melting animals into goo. The obvious hypothesis: maybe the susceptible species just don’t have enough immune genes. Hannah put that idea to the test, combing through annotated sea star genomes and sorting immune genes into functional categories. The plot twist? Gene count isn’t the answer. Susceptible and resistant species carry similar immune gene arsenals — so the real story is about how those genes are switched on and off, not how many there are. One clean result, and suddenly the whole field is asking a better question.

Download Hannah’s capstone presentation (PDF) · Read Hannah’s capstone paper (PDF) · GitHub repo

Noah Ozguner — Do oysters get stressed when you shake them? 🦪

Oyster farmers “tumble” their oysters to sculpt prettier shells — but is it secretly stressing the animals out? Noah grabbed two sizes of oyster seed, gave them a 20-minute shake, and tracked their metabolism for a week with a color-changing assay. Tiny oysters? Totally unbothered. But the big ones threw a delayed metabolic curveball — no reaction at first, then a dramatic spike a full week later (p = 0.0006). Turns out bigger oysters carry a hidden energetic hangover, and it shows up on a timer.

Download Noah’s poster (PDF)

Hazel Abrahamson-Amerine — What a heatwave does to an oyster’s fuel tank 🔥

Oysters bank energy as glycogen — their cellular savings account. Hazel asked what happens to that account during a marine heatwave. Using a glow-in-the-dark glycogen assay, they cranked oysters to 32 °C for 48 hours and watched the reserves crash. The hopeful part: give them a break from the heat, and they slowly rebuild the stash. Recovery is real — it just takes time. Next up: where in the body it happens, and whether oysters toughen up for round two.

Download Hazel’s presentation (PDF)

Christina Zhang — Can oysters be trained to take the heat? 🌡️

If a heatwave hits twice, does surviving the first one help with the second? Christina tested whether oysters can be “primed” — toughened up by an earlier non-lethal heat exposure. She gave Pacific oyster seed different thermal histories (cool or hot, then cool or hot again) and read out their metabolism across a temperature gradient. The verdict: priming changes the game. Below 35 °C, primed oysters ran a calmer, lower metabolic response — a sign of tolerance — while unprimed oysters spiked. But push past 37 °C and the stress response overrides any priming benefit, and mortality climbs. Oysters can learn to take the heat, but only up to a point.

Download Christina’s poster (PDF)

Four students, four big swings, four home runs. Congrats Hannah, Noah, Hazel, and Christina — the lab is lucky to have you. 🎉

Why Standard Tools Fall Short

If you work in genomics, you are probably familiar with tools like DESeq2 and edgeR for differential gene expression analysis. These are workhorses of the field, but they are built for pairwise comparisons: treatment vs. control, timepoint A vs. timepoint B. Our coral data has a richer structure — expression measured across multiple genes, multiple species, and multiple samples over time — that these tools cannot natively capture.

We needed methods that could discover coordinated expression patterns spanning all three of those dimensions simultaneously.

Barnacle: Sparse Tensor Decomposition for Multi-Species Expression

Enter Barnacle, a sparse tensor decomposition tool originally developed for marine metatranscriptomics. The idea is elegant: instead of flattening your data into a two-dimensional matrix, you represent it as a 3D tensor (Gene x Taxon x Sample) and decompose it into a small number of sparse components. Each component captures a distinct co-expression program — a group of genes that behave similarly across species and conditions.

The critical tuning decisions are:

• How many components (rank)? Too few and you miss real signals; too many and you start fitting noise. We use a cross-validation strategy — fit on half the data, test reconstruction on the held-back half — to find the sweet spot.
• How sparse? Sparsity controls which genes are assigned to each component. We use split-half reproducibility: if both halves of a random split identify the same core genes for a component, the sparsity threshold is keeping real members.

Applied to our E5 coral timeseries data, Barnacle revealed biologically meaningful components. One component, for example, aligned closely with known calcification genes — a result that would be difficult to extract from traditional pairwise analyses.

With these components in hand, we now have groups of co-regulated genes that we can interrogate for epigenetic regulatory patterns and use to test hypotheses about how environmental stress propagates through molecular networks.

The Epigenetic Layer: More Than One Regulator

Epigenetics — heritable changes in gene activity that do not alter the DNA sequence itself — is central to our project. In marine invertebrates, the epigenetic landscape includes:

• DNA methylation — In contrast to mammals, marine invertebrate methylation is mosaic and concentrated in gene bodies rather than promoters.
• Long non-coding RNA (lncRNA) — Regulatory RNA molecules numbering in the tens of thousands.
• MicroRNA (miRNA) — Short regulatory RNAs that target messenger RNA for degradation or translational repression.

Each of these layers can influence gene expression, and they likely interact with each other. Simple pairwise correlations between a single epigenetic mark and expression only tell part of the story. We wanted a model that could integrate all three layers and identify which features — and which combinations — actually drive expression.

Elastic Net Regression: Predicting Expression from Epigenetics

This is where Elastic Net regression comes in. Elastic Net is a regularized regression method that blends two complementary penalties:

• Lasso (the “harsh editor”) drives coefficients to exactly zero, performing feature selection by removing irrelevant predictors.
• Ridge (the “compromiser”) shrinks correlated predictors together rather than arbitrarily choosing one, handling multicollinearity gracefully.

Our predictor space is formidable: roughly 50 miRNAs, 10,000 lncRNAs, and 20,000 gene-body methylation features, all predicting expression across approximately 40 samples per species. This is a classic p >> n problem, and Elastic Net is well suited for it.

We trained models using 80/20 train/test splits repeated across 50 replicates, validated predictions using noise injection (if the model is real, adding noise to predictors should degrade performance), and applied stability selection (Faletto and Bien 2022) to focus on features that are consistently selected across resampled datasets rather than artifacts of any single split.

Video

What We Found

The results are striking and represent, to our knowledge, the first integrated picture of multi-layer epigenetic regulation of gene expression in a marine invertebrate system:

• Per-gene regulatory profiles reveal that individual genes differ substantially in which epigenetic features predict their expression. Some are primarily methylation-driven, others lncRNA-driven, and many are regulated by a combination.
• High-level type compositions show systematic patterns in how regulatory strategies are distributed across the genome.
• Regulatory network maps capture how epigenetic features interact, suggesting coordinated regulatory programs rather than independent, isolated effects.

Looking Ahead

This work is a starting point. The tensor decomposition framework can be extended to incorporate additional data types (e.g., lncRNA expression, genes found in fewer than all species), and the elastic net models can be refined as more samples and species are added. The broader goal — understanding how environmental stress rewires epigenetic regulation and whether those changes can be inherited — remains one of the most important questions in coral biology and climate adaptation.

We are grateful to the eScience Incubator Program and particularly to Vaughn Iverson for pushing us to think about our data in new ways. The combination of domain expertise in marine biology with data science methods has opened analytical doors that neither community would have found alone.

For more information, visit the E5 project and Barnacle documentation.