Study Questions Effectiveness of Foundation Models for Genomics Due to High Entropy in DNA Sequences
Researchers have identified high sequence entropy as a core reason why foundation models trained on genomic DNA underperform compared to those trained on natural language. By training ensembles of models on both text and DNA, they found that genomic data produces near-uniform output distributions, model disagreement, and unstable embeddings. The findings call into question whether self-supervised sequence training — the dominant paradigm for large language models — is appropriate for genomic data at all.
A study accepted to the LMLR Workshop at ICLR 2026 investigates why foundation models in genomics have shown mixed results relative to their natural language processing counterparts. The researchers trained ensembles of models on both text and DNA sequences, then analyzed predictions, static embeddings, and empirical Fisher information flow across these models. Their central finding is that genomic sequences exhibit high entropy from the perspective of unseen token prediction, causing models to produce near-uniform output distributions and disagree substantially with one another even when matched in architecture, training procedure, and data. Additionally, models trained on DNA appear to concentrate Fisher information in embedding layers rather than learning inter-token relationships, suggesting they fail to capture meaningful sequential structure. These results challenge the foundational assumption that self-supervised training on raw sequences — which has proven powerful in NLP — transfers effectively to genomic contexts. The authors argue that alternative training strategies or additional data modalities may be necessary for genomic foundation models to achieve robust, generalizable capabilities.
What's missing
The study does not address whether incorporating multimodal biological data (e.g., epigenomic, proteomic, or functional annotation data) could mitigate the entropy problem. The workshop paper format also means findings have not yet undergone full peer review.
What different sources said
- arXiv cs.LGCenter
Entropy, Disagreement, and the Limits of Foundation Models in Genomics
Related
Gut Bacteria Enzyme Found to Break Down Heat-Processed Food Compounds, Producing Novel Biogenic Amines
Researchers have discovered that an enzyme in common gut bacteria can degrade N-epsilon-carboxymethyllysine (CML), a compound formed during thermal food processing, producing previously unknown biogenic amines. The enzyme, ornithine decarboxylase SpeC from enterobacteria, acts on CML and related modified lysine derivatives through a low-level 'underground' catalytic activity. This finding suggests a previously unrecognized communication axis between thermally processed dietary compounds and gut microbial physiology, with potential implications for host health.
Full-Length Gene Sequencing Reveals Two Distinct Bacterial Communities in Black-Legged Ticks Expanding Into Canada
Researchers used Oxford Nanopore full-length 16S rRNA gene sequencing to characterize the microbiome of Ixodes scapularis black-legged ticks collected in Nova Scotia, Canada, distinguishing between tick-adapted bacteria and environmentally acquired bacteria. The study comes as I. scapularis — the primary vector of Lyme disease — is rapidly expanding northward into Canada due to climate change. The findings suggest that environmentally derived bacteria in tick microbiomes are not mere contamination, which has implications for how tick microbiome data is collected and interpreted across surveillance studies.
Study Identifies Metabolic Link Between Cell Envelope Stress and Biofilm Formation in Bacteria
Researchers have discovered that the metabolite acetyl-CoA directly inhibits enzymes that degrade the bacterial signaling molecule c-di-GMP, connecting cell envelope biosynthesis stress to biofilm formation in Pseudomonas aeruginosa. The study found that sub-inhibitory concentrations of antibiotics targeting early peptidoglycan biosynthesis — but not other antibiotic classes — elevate c-di-GMP levels by reducing phosphodiesterase activity, with acetyl-CoA competing for the enzyme active site. Because the relevant enzyme domain is broadly conserved across bacterial species, this checkpoint mechanism may be widespread and could have implications for understanding antibiotic-induced biofilm responses.