Run ProteinMPNN online

Pick ProteinMPNN when you already have a backbone and need candidate sequences. For de novo backbone generation, use RFantibody, BindCraft, or BoltzGen first and feed the output PDB here.

ProteinMPNN (Dauparas et al., Science 2022). A message-passing graph neural network that scores the 20 canonical residues at every backbone position, conditioned on Cα / backbone coordinates. Sampling at sampling_temp produces candidate sequences that fold into the input geometry.

When it fits:

• You already have a backbone and need candidate sequences for it.
• You want to redesign a binder produced by RFdiffusion, RFantibody, BindCraft, or BoltzGen.
• You want to thread alternative sequences through a curated PDB before ordering.

Use the score legend below to read results. Each tool reports a subset of these depending on whether it does design, sequence recovery, or structure prediction.

ipTM

Predicted confidence in the binder to target interface. Higher is better. Aim above roughly 0.7 on a tractable target.

pLDDT

Per-residue confidence in the predicted fold. Higher means the model is more sure of that part of the structure.

i_pAE and pAE

Predicted alignment error, at the interface (i_pAE) or across the whole structure (pAE). Lower is better.

ProteinMPNN recovery

Fraction of native residues recovered when ProteinMPNN redesigns a known sequence on its native backbone. Higher is better; well calibrated above roughly 0.4 on diverse folds.

1 min per run on a dedicated GPU. Billing is by the second of compute, so a faster preset costs less.

We keep a plain-English overview of ProteinMPNN on the main Ranomics site — what the model does under the hood, the kinds of targets it works on, and where it fits inside a full design-to-wet-lab campaign.

If you are picking between ProteinMPNN and a sibling algorithm, these run on the same hub against the same target.

Run AlphaFold2 online

Pick AF2 when you need the gold-standard structure prediction with calibrated pLDDT + PAE. For faster single-sequence folds use ESMFold (D4); for affinity-aware folds use Boltz-2 (D6).

Run ColabFold online

Pick ColabFold when you need a fast no-MSA fold — 1-2 min per run, no MMseqs2 round-trip. Pair with AF2 standalone (D2) when you want full MSA + templates, or with ESMFold (D4) for single-sequence monomers on an even smaller GPU.

Run ESMFold online

Pick ESMFold when you need the fastest possible monomer fold - no MSA, no multimer, single-sequence ESM-2 language-model prediction. Pair with ColabFold (D3) for multimers or AF2 standalone (D2) for full MSA-backed accuracy.

Can I run ProteinMPNN online without a local GPU?

Yes. Upload a backbone PDB and Ranomics Tools runs ProteinMPNN on a dedicated GPU in seconds. You get ranked sequence redesigns plus per-position recovery — no install, no CUDA.

How much does one ProteinMPNN job cost?

Billing is by the second. A typical ProteinMPNN job costs a fraction of a cent because the model finishes in under a minute on most backbones. New accounts start with a $5 wallet balance, which is enough for thousands of runs.

Does ProteinMPNN need an MSA?

No. ProteinMPNN is a structure-conditioned sequence model — it takes the backbone coordinates directly and never queries an MSA. That makes it the fastest route from a designed backbone to a sequence you can synthesise.

Dauparas et al., Science 2022