Literature · wanglab
BioReason-Pro SFT
Multimodal reasoning model for protein function prediction with structured chain-of-thought rationales.
What BioReason-Pro SFT is
BioReason-Pro SFT is a supervised fine-tuned biological language model developed by the Wang Lab. It addresses the task of protein function prediction by conducting multimodal biological reasoning across protein sequences, genomic variants, and pathway information.
Rather than generating direct label classifications without context, the model produces structured chain-of-thought rationales alongside functional annotations. This provides researchers with interpretable justifications derived from molecular representations and biological interaction data.
How it works
The model uses a Qwen3-4B language model backbone integrated with ESM3 protein embeddings and a Gene Ontology (GO) graph encoder. It also incorporates contextual biological data from InterPro functional domains and STRING protein-protein interaction networks.
The model is supervised fine-tuned on the wanglab/bioreason-pro-sft-reasoning-data dataset, allowing it to take textual prompts containing biological information and output structured reasoning traces with functional predictions.
Key facts
- Backbone architecture
- Qwen3-4B
- Protein embeddings
- ESM3
- Graph representation
- Gene Ontology (GO) graph encoder
- Biological context sources
- InterPro domains, STRING interactions
- Training dataset
- wanglab/bioreason-pro-sft-reasoning-data
- GPU memory requirement
- 24 GB
What you can use it for
- Protein function prediction
- Annotates protein sequences with functional labels and Gene Ontology terms using sequence embeddings and network context.
- Chain-of-thought biological reasoning
- Generates explicit reasoning paths that connect protein features and pathway data to functional conclusions.
- Interpreting genomic and pathway context
- Assesses biological mechanisms by reasoning over InterPro domains, STRING interactions, and genomic variants.
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
BioReason-Pro SFT
Strengths and limits
Good at
- Integrates multimodal features including ESM3 embeddings, Gene Ontology graph structures, and STRING interactions into a single reasoning model.
- Produces transparent, structured chain-of-thought rationales alongside functional predictions.
- Operates within a standard 24 GB GPU memory footprint.
Watch out for
- Requires a system with at least 24 GB of GPU memory to execute.
- Relies on the availability and accuracy of external biological context such as InterPro domains and STRING network interactions.
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 BioReason-Pro SFT 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 BioReason-Pro SFT 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 backbone architecture powers BioReason-Pro SFT?
It is built upon a Qwen3-4B foundation model, coupled with ESM3 protein embeddings and a Gene Ontology graph encoder.
What types of inputs and outputs does the model process?
The model takes biological prompts as text input and generates text completions containing structured reasoning traces and functional annotations.
How does this model differ from BioReason-Pro RL?
BioReason-Pro SFT is the supervised fine-tuned checkpoint, whereas BioReason-Pro RL is a distinct checkpoint further refined through reinforcement learning.
How to cite it
Fallahpour, A., Seyed-Ahmadi, A., Idehpour, P., Ibrahim, O., Gupta, P., Naimer, J., Zhu, K., Shah, A., Ma, S., Adduri, A., Güloglu, T., Liu, N., Cui, H., Jain, A., de Castro, M., Fallahpour, A., Cembellin-Prieto, A., Stiles, J. S., Nemčko, F., Nevue, A. A., Moon, H. C., Sosnick, L., Markham, O., Duan, H., Lee, M. Y. Y., Salvador, A. F. M., Maddison, C. J., Thaiss, C. A., Ricci-Tam, C., Plosky, B. S., Burke, D. P., Hsu, P. D., Goodarzi, H., & Wang, B. (2026). BioReason-Pro: Advancing Protein Function Prediction with Multimodal Biological Reasoning. bioRxiv, 2026.03.19.712954. doi:10.64898/2026.03.19.712954
Written from BioReason-Pro SFT'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.