The genome is a sequence that encodes the DNA, RNA, and proteins that orchestrate an organism’s function. We present Evo, a long-context genomic foundation model with a frontier architecture trained on millions of prokaryotic and phage genomes, and report scaling laws on DNA to complement observations in language and vision. Evo generalizes across DNA, RNA, and proteins, enabling zero-shot function prediction competitive with domain-specific language models and the generation of functional CRISPR-Cas and transposon systems, representing the first examples of protein-RNA and protein-DNA codesign with a language model. Evo also learns how small mutations affect whole-organism fitness and generates megabase-scale sequences with plausible genomic architecture. These prediction and generation capabilities span molecular to genomic scales of complexity, advancing our understanding and control of biology.

The rise of digital data and computing power have contributed to significant advancements in artificial intelligence (AI), leading to the use of classification and prediction models in health care to enhance clinical decision-making for diagnosis, treatment and prognosis. However, such advances are limited by the lack of reporting standards for the data used to develop those models, the model architecture, and the model evaluation and validation processes. Here, we present MINIMAR (MINimum Information for Medical AI Reporting), a proposal describing the minimum information necessary to understand intended predictions, target populations, and hidden biases, and the ability to generalize these emerging technologies. We call for a standard to accurately and responsibly report on AI in health care. This will facilitate the design and implementation of these models and promote the development and use of associated clinical decision support tools, as well as manage concerns regarding accuracy and bias.

All of life encodes information with DNA. Although tools for genome sequencing, synthesis and editing have transformed biological research, we still lack sufficient understanding of the immense complexity encoded by genomes to predict the effects of many classes of genomic changes or to intelligently compose new biological systems. Artificial intelligence models that learn information from genomic sequences across diverse organisms have increasingly advanced prediction and design capabilities^1,2. Here we introduce Evo 2, a biological foundation model trained on 9 trillion DNA base pairs from a highly curated genomic atlas spanning all domains of life to have a 1 million token context window with single-nucleotide resolution. Evo 2 learns to accurately predict the functional impacts of genetic variation—from noncoding pathogenic mutations to clinically significant BRCA1 variants—without task-specific fine-tuning. Mechanistic interpretability analyses reveal that Evo 2 learns representations associated with biological features, including exon–intron boundaries, transcription factor binding sites, protein structural elements and prophage genomic regions. The generative abilities of Evo 2 produce mitochondrial, prokaryotic and eukaryotic sequences at genome scale with greater naturalness and coherence than previous methods. Evo 2 also generates experimentally validated chromatin accessibility patterns when guided by predictive models^3,4 and inference-time search. We have made Evo 2 fully open, including model parameters, training code⁵, inference code and the OpenGenome2 dataset, to accelerate the exploration and design of biological complexity.

There is increased interest in nonpharmacological treatments to reduce pain after total knee arthroplasty. Yet, little consensus supports the effectiveness of these interventions. To systematically review and meta-analyze evidence of nonpharmacological interventions for postoperative pain management after total knee arthroplasty. Database searches of MEDLINE (PubMed), EMBASE (OVID), Cochrane Central Register of Controlled Trials (CENTRAL), Cochrane Database of Systematic Reviews, Web of Science (ISI database), Physiotherapy Evidence (PEDRO) database, and ClinicalTrials.gov for the period between January 1946 and April 2016. Randomized clinical trials comparing nonpharmacological interventions with other interventions in combination with standard care were included. Two reviewers independently extracted the data from selected articles using a standardized form and assessed the risk of bias. A random-effects model was used for the analyses. Postoperative pain and consumption of opioids and analgesics. Of 5509 studies, 39 randomized clinical trials were included in the meta-analysis (2391 patients). The most commonly performed interventions included continuous passive motion, preoperative exercise, cryotherapy, electrotherapy, and acupuncture. Moderate-certainty evidence showed that electrotherapy reduced the use of opioids (mean difference, -3.50; 95% CI, -5.90 to -1.10 morphine equivalents in milligrams per kilogram per 48 hours; P = .004; I2 = 17%) and that acupuncture delayed opioid use (mean difference, 46.17; 95% CI, 20.84 to 71.50 minutes to the first patient-controlled analgesia; P < .001; I2 = 19%). There was low-certainty evidence that acupuncture improved pain (mean difference, -1.14; 95% CI, -1.90 to -0.38 on a visual analog scale at 2 days; P = .003; I2 = 0%). Very low-certainty evidence showed that cryotherapy was associated with a reduction in opioid consumption (mean difference, -0.13; 95% CI, -0.26 to -0.01 morphine equivalents in milligrams per kilogram per 48 hours; P = .03; I2 = 86%) and in pain improvement (mean difference, -0.51; 95% CI, -1.00 to -0.02 on the visual analog scale; P < .05; I2 = 62%). Low-certainty or very low-certainty evidence showed that continuous passive motion and preoperative exercise had no pain improvement and reduction in opioid consumption: for continuous passive motion, the mean differences were -0.05 (95% CI, -0.35 to 0.25) on the visual analog scale (P = .74; I2 = 52%) and 6.58 (95% CI, -6.33 to 19.49) opioid consumption at 1 and 2 weeks (P = .32, I2 = 87%), and for preoperative exercise, the mean difference was -0.14 (95% CI, -1.11 to 0.84) on the Western Ontario and McMaster Universities Arthritis Index Scale (P = .78, I2 = 65%). In this meta-analysis, electrotherapy and acupuncture after total knee arthroplasty were associated with reduced and delayed opioid consumption.