Genomic language models could transform medicine but not yet
Bibliographic record
Abstract
In February 2025, researchers announced Evo2, a genome language model (gLM) trained on over 128,000 genomes, encompassing over 9.3 trillion DNA base pairs 1 . This computational scale matches leading text-based LLMs, representing a significant milestone for genomic AI 2 . Unlike protein language models, which train to understand the 2% of human DNA that is encoded into amino acids and folded into proteins, gLMs train to understand the entire genome 3 . This largely consists of understanding the role of the remaining 98% of human DNA that is non-coding. Non-coding DNA contains crucial regulatory elements that coordinate gene expression across different cell types and developmental stages 4 , and the precise mechanisms governing this regulation are increasingly being unraveled. This field of study is known as regulatory genomics 4 , and gLMs have emerged as promising tools to study it. The introduction of Evo2 represents both important progress for the field and highlights critical questions about what these models learn and how they might be applied. This article examines gLMs in the context of Evo2, highlighting their potential for biological research and medicine while exploring the technical barriers and ethical challenges—from data privacy to dual-use risks—that will shape their clinical future.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.009 | 0.033 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.001 | 0.005 |
| Scholarly communication | 0.006 | 0.011 |
| Open science | 0.002 | 0.004 |
| Research integrity | 0.003 | 0.007 |
| Insufficient payload (model declined to judge) | 0.012 | 0.006 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".