Literature · ZhejiangLab
OneGenome-Rice
A genomic foundation model analysing long DNA sequences for rice functional genomics and precision breeding.
What OneGenome-Rice is
OneGenome-Rice (OGR) is a generative genomic foundation model developed by ZhejiangLab for functional genomics and precision breeding in rice (Oryza sativa). Built on a Mixture-of-Experts architecture, it is designed to process and interpret long genomic sequences across diverse rice lineages.
The model addresses essential agricultural and biological tasks, including genetic variant analysis, gene expression prediction, and the modelling of traits relevant to crop breeding. By learning patterns across both cultivated varieties and wild relatives, it serves as a computational resource for studying rice genomics.
How it works
The model uses a Mixtral-style Mixture-of-Experts (MoE) architecture comprising 12 layers, 1.25 billion total parameters, and 0.33 billion activated parameters per token. It contains 8 experts with an expert hidden dimension of 4,096, routing each token to 2 experts. The attention backbone operates with a hidden dimension of 1,024 across 16 attention heads using grouped-query attention with 8 key-value groups, and a padded vocabulary size of 128.
OneGenome-Rice is capable of handling long-range sequence context up to 1 million base pairs (1 Mb). It was pre-trained on a curated genomic corpus comprising 422 rice genomes spanning cultivated varieties and wild Oryza diversity.
Key facts
- Architecture
- Mixture-of-Experts (MoE, Mixtral)
- Total parameters
- 1.25B
- Activated parameters
- 0.33B
- Number of layers
- 12
- Number of experts
- 8 (2 selected per token)
- Training data
- 422 cultivated and wild Oryza genomes
What you can use it for
- Genomic variant analysis
- Assessing the effects and distribution of sequence variations across rice lineages.
- Gene expression prediction
- Predicting transcriptional activity and expression patterns from extended genomic DNA sequences.
- Breeding trait modelling
- Modelling phenotypes and traits of interest to support precision crop breeding programmes.
Run it here
No installation and no GPU of your own: fill in the inputs and it runs on our servers. Results stay in your browser unless you sign in.
Runs on On-demand GPU
Runs on our servers with on-demand compute. A first run needs time to load the model; active capacity can be reused and scales down when idle.Run this model
OneGenome-Rice
Strengths and limits
Good at
- Processes long-range sequence contexts of up to 1 million base pairs.
- Uses a sparse Mixture-of-Experts design that activates only 0.33 billion parameters during execution.
- Pre-trained on a diverse genomic collection spanning 422 cultivated and wild Oryza accessions.
Watch out for
- Specialised for rice (Oryza) species, limiting direct application to unrelated plant taxa.
- Standard deployments configure the model as a text-generation pipeline, requiring validation before relying on downstream outputs.
Inputs and outputs
You provide
Prompt
The model continues from here.
You get back
The model's completion.
Licence
Check the licence before commercial use · See model card
The See model card licence carries terms worth reading before you depend on it.
Questions
Can I use OneGenome-Rice commercially?
Check the licence before commercial use. The See model card licence carries terms worth reading before you depend on it.
How do I run OneGenome-Rice without installing anything?
Use the form under “Run it here” on this page: fill in the inputs and press Run model. It runs on our servers on 1× A100 16GB, typically in 12s, and a run costs about 70 credits. Visitors without an account get 1,000 credits a month.
What species does OneGenome-Rice cover?
The model is pre-trained on a curated collection of 422 genomes covering cultivated rice (Oryza sativa) as well as wild Oryza diversity.
What is the maximum sequence length the model can accept?
The model can process DNA sequences up to 1 million base pairs (1 Mb) in length.
How many parameters are engaged during inference?
Although the model contains 1.25 billion total parameters, its Mixture-of-Experts architecture activates only 0.33 billion parameters per token across 2 selected experts.
Written from OneGenome-Rice's own documentation (model card). Statements citing a figure that does not appear there are removed before publication. Licence and running details come from our registry.