Know which mutations will work before you spend weeks in the lab.

What Is NeuroAutomata

NeuroAutomata is a web-based protein mutation scoring tool built on ESM-2, a protein language model developed by Meta AI and trained on millions of protein sequences from across the tree of life. Think of it as a second opinion on your mutagenesis experiments — one that's read every protein sequence in UniProt.

NeuroAutomata now also supports ESMC, a protein language model family released under an open MIT license (commercial use included) by the Chan Zuckerberg Biohub — the nonprofit research institute the original ESM team joined in late 2025. Because it's openly licensed, we can offer it to you directly.

You can choose the model that fits your work:

You paste a sequence, and NeuroAutomata predicts which mutations are likely tolerable and which ones will probably break your protein. It works on any protein from any organism — even sequences that have never been experimentally characterized — because its predictions come from evolutionary patterns across all known protein families, not from labeled training data specific to your protein.

It runs in your browser. There's nothing to install. Every result you generate — charts and raw scores — is downloadable. Your sequences are processed in memory and deleted from our systems within an hour. Full details in our privacy policy.

NeuroAutomata is built for protein engineers, biochemists, and anyone running mutagenesis on their own proteins. It is free to start, with no card required. Run it on your sequences and tell us what works and what we should fix.

It Already Knew About EGFP

See It in Action

We loaded 9 proteins — fluorescent reporters, therapeutic targets, industrial enzymes, and a CRISPR domain — and scored all of them. Before running any mutations, NeuroAutomata shows you how your proteins relate to each other.

How It Works

What It Found in GFP

The underlying model was trained on millions of protein sequences from across the tree of life. It learned which positions matter and which are flexible — without being told anything about structure, function, or experimental data. When we scored GFP:

Nobody told the model about the GFP chromophore. Nobody provided GFP-specific training data. It figured this out from sequence patterns alone, in 30 seconds.

Validated Against Real Experiments

We checked NeuroAutomata's predictions against published experimental data: deep mutational scanning results from ProteinGym, a reference benchmark where researchers made the mutations in the lab and measured what each one did. These are clinically and pharmacologically real proteins, not toy examples. They include cancer-associated tumor suppressors, drug-metabolizing and drug-dosing enzymes, and a cell-surface receptor.

Across these seven measurements, the median rank correlation is 0.5298. For scale, ESM-2 650M, NeuroAutomata's default model, averages 0.414 across the full ProteinGym benchmark of 217 assays (leaderboard CSV, accessed 2026-05-08). Most of this cohort lands above that benchmark-wide average, though not all of it does.

Where the model is weakest is binding. The two lowest scores are both binding measurements: the insulin receptor at 0.3989, and calmodulin, a calcium-sensing protein we also benchmarked, at 0.212, which is well below the model's own 217-assay average. We report this alongside the strong results rather than burying it. The in-app Learn section walks through the full method: what it does well, and where it falls short.

What Happens to Your Sequences

Your sequences are processed in memory, held in a temporary queue for up to one hour so you can download your results, and then automatically deleted. Nothing is saved to disk. We don't train models on your data. Everything you generate is downloadable, so keep the results you want.

Some processing runs on external services. Running your sequence through the model (embeddings, mutation scoring, and landscape scans) happens on a contracted GPU provider that acts as our data processor. Structure prediction sends your sequence to a separate third-party public API that we don't control. Our privacy policy names each provider and explains how your data is handled.

What You Get

• Works in your browser — nothing to install, nothing to configure
• Results in seconds — not days, not hours
• Works on any protein — from any organism, even ones never studied before
• Download everything you generate — charts and raw data as CSV. Your results, your files.

Use It in Claude and ChatGPT

Run NeuroAutomata where you already work. Add it as a custom connector (currently a beta feature in both assistants) with one URL: works on every Claude plan, and on paid ChatGPT plans with Developer mode turned on. Read the guide →