Protein Design · Harvard / Broad Institute
EVOLVEpro
Active learning model for directed evolution that proposes mutant sequences from small experimental batches.
What EVOLVEpro is
EVOLVEpro is an active learning tool developed by Harvard and the Broad Institute for protein engineering and directed evolution. It is designed to iteratively guide wet-lab experiments by taking incoming experimental assay results and proposing the next round of candidate mutants.
By learning continuously from newly acquired assay data, EVOLVEpro shifts directed evolution from large-scale library screening towards a computational design process. The system has been validated across diverse enzymes using as few as 24 experimental measurements per round.
How it works
EVOLVEpro is built on top of the ESM2 protein language model. It takes a wild-type protein sequence, experimental assay measurements supplied in CSV format, and a specified batch size. Using a few-shot active learning framework, it processes the experimental results to propose and rank the next set of candidate variants according to predicted fitness and sequence diversity.
Key facts
- Version
- 1.0
- Base architecture
- ESM2
- Domain
- Protein design
- Tasks
- Directed evolution, active learning
- Required GPU memory
- 16 GB
What you can use it for
- Enzyme engineering
- Iteratively optimising diverse enzymes by evaluating experimental measurements and generating targeted variant batches.
- Low-throughput directed evolution
- Designing subsequent testing batches when experimental screening throughput is limited to as few as 24 mutants per round.
- Diversity-aware sequence proposal
- Generating ranked candidate lists that balance predicted functional fitness with sequence diversity based on prior assay outcomes.
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.
Not runnable yet
This model isn't available to run right now.Run this model
EVOLVEpro
Strengths and limits
Good at
- Functions effectively in few-shot active learning settings with as few as 24 measurements per round.
- Builds directly upon the ESM2 protein language model representation.
- Supplies candidate recommendations balanced across predicted fitness and sequence diversity.
Watch out for
- Requires empirical assay measurements in CSV format to guide successive rounds of design.
- Demands a minimum of 16 GB of GPU memory.
Inputs and outputs
You provide
Wild-type sequence
Single-letter amino acid code. Whitespace and FASTA headers are stripped automatically.
Assay data (CSV)optional
Prior rounds of measurements. Leave empty to seed round one.
Next batch sizeoptional
You get back
Ranked next-round candidates with predicted fitness and diversity.
Licence
Commercial use permitted · MIT
The MIT licence allows use in commercial research and products.
Questions
Can I use EVOLVEpro commercially?
Commercial use permitted. The MIT licence allows use in commercial research and products.
How do I run EVOLVEpro 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 30s, and a run costs about 210 credits. Visitors without an account get 1,000 credits a month.
What inputs are required to run EVOLVEpro?
The model requires a wild-type protein sequence, experimental assay data formatted as a CSV, and the desired batch size for the next round of proposals.
What does the model produce as output?
It generates a table of ranked candidate mutants along with their predicted fitness and sequence diversity metrics.
What base model does EVOLVEpro use?
EVOLVEpro is built on top of ESM2.
What is the minimum number of measurements needed per round?
The model has been validated using as few as 24 measurements per round.
Written from EVOLVEpro's own documentation. Statements citing a figure that does not appear there are removed before publication. Licence and running details come from our registry.