Upload your CSV. Get a ranked dashboard across three kinetic channels. Compare scoring methods side by side. Know exactly which mutations to run in the lab next.
Every mutation is scored across three independent effect channels, not collapsed into a single misleading number.
Detects mutations that impede the folding pathway. The protein may eventually fold, but the kinetic cost matters for throughput and yield.
Quantifies thermodynamic instability in the folded state. Catches mutations that weaken the native structure without necessarily blocking folding.
Flags mutations that increase the rate of unfolding. A protein that folds fine but unfolds too fast is functionally compromised.
| Mutation | Rank | Fold Slow | Destab | Unfold Acc | Flag |
|---|---|---|---|---|---|
| G148R | #1 | 0.91 | 0.76 | 0.97 | seq-only |
| A203V | #2 | 0.72 | 0.91 | 0.48 | ok |
| L57P | #3 | 0.95 | 0.38 | 0.64 | seq-only |
| D312N | #4 | 0.44 | 0.69 | 0.81 | ok |
Drop in hundreds of mutations with optional protein metadata. Validated on ingest, ranked on output.
Export ranked results as PDF or CSV. Share with collaborators or attach to grant applications.
Every formula is modular and exposed. Update weights, swap methods, audit the math. No black boxes.
Rows where sequence-only scoring may be unreliable are flagged automatically. Know when to reach for structural data.
The entire scoring engine is exportable to GitHub. Reproduce, fork, extend. Full scientific transparency.
Designed for ESM, ProteinMPNN, AlphaFold/ColabFold, UniProt, and PDB inputs. The architecture is ready.
See how different scoring approaches rank the same mutation. Side by side. No switching tools.
Sequence-neighbor context around the mutation site
3D spatial relationships and contact networks
Conservation and covariation from MSA data
Protein language model embeddings and predictions
MutaRank doesn't predict universal folding. It tells you where to look next, and which of your models to trust.