Genomics & Single-Cell · DeepMind
Enformer
Predicts gene expression and regulatory tracks across long genomic windows to assess variant effects.
What Enformer is
Enformer is a genomics model developed by DeepMind that predicts gene expression directly from nucleotide sequence. It is designed to model regulatory influence over broad genomic distances, addressing the challenge of mapping how distal regulatory elements modulate gene activity.
By operating over a 200kb window, the model accounts for long-range enhancer–promoter interactions up to 100kb away. This makes it an established tool for assessing the impact of non-coding genetic variants on transcriptional regulation and chromatin features.
How it works
The model accepts genomic nucleotide sequences or variant sequences alongside an organism specification, operating across a 200kb receptive field. It outputs predictions across genomic bins for functional tracks, including CAGE, DNase sensitivity, and histone marks.
Key facts
- Version
- 1.0
- Domain
- Genomics
- Input modalities
- Nucleotide sequence
- Context window
- 200kb
- Required GPU memory
- 24 GB
What you can use it for
- Variant effect prediction
- Evaluate how non-coding genomic variants influence expression profiles by comparing reference and alternative sequences.
- Long-range enhancer modelling
- Capture regulatory influences and interactions between distal enhancers and target promoters situated up to 100kb away.
- Regulatory track prediction
- Generate binned genomic predictions for CAGE, DNase sensitivity, and histone modifications directly from sequence.
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
Enformer
Strengths and limits
Good at
- Incorporates a 200kb receptive field to represent regulatory elements located up to 100kb away.
- Provides simultaneous binned predictions for multiple regulatory modalities, including CAGE, DNase, and histone marks.
- Directly evaluates non-coding variant effects from sequence inputs.
Watch out for
- Demands substantial hardware, requiring at least 24 GB of GPU memory.
- Cannot evaluate interactions originating beyond its 200kb sequence window.
Inputs and outputs
You provide
Genomic sequence
Up to 196,608 bp centred on the region of interest. Shorter is padded, and the result says how much.
Variant sequenceoptional
The same window carrying the variant. Given both, the result is the difference.
Organismoptional
You get back
CAGE, DNase and histone mark predictions per bin.
Licence
Commercial use permitted · Apache-2.0
The Apache-2.0 licence allows use in commercial research and products.
Questions
Can I use Enformer commercially?
Commercial use permitted. The Apache-2.0 licence allows use in commercial research and products.
How do I run Enformer 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 40GB, typically in 25s, and a run costs about 70 credits. Visitors without an account get 1,000 credits a month.
What inputs does Enformer require?
It accepts genomic sequence, variant sequence, and an organism selection.
What are the model's outputs?
The model outputs tabular predictions across genomic bins for CAGE, DNase, and histone marks.
What window size does the model cover?
It processes genomic sequences across a 200kb window, enabling the detection of interactions up to 100kb away.
What are the computational requirements to run Enformer?
The model requires 24 GB of GPU memory.
Written from Enformer's own documentation. Statements citing a figure that does not appear there are removed before publication. Licence and running details come from our registry.