SpotONE
Hot SPOTs ON protein complexes with Extremely randomized trees: sequence-only prediction
0.82
Accuracy on independent test set
0.83
AUROC
0.91
Precision
0.85
F1-score
Hot spots (HS), the small subset of interface residues that contribute disproportionately to binding free energy, are prime targets for drug design and the rational engineering of protein–protein interactions. Identifying them experimentally is costly and low-throughput; computational prediction from sequence alone offers a scalable alternative.
SpotONE is a machine-learning predictor built on Extremely Randomized Trees that classifies every residue in a submitted protein sequence as a hot spot or non-hot spot using sequence-derived features exclusively. No structural data is required, making it immediately applicable to any protein with a known sequence, and results are delivered directly to the user’s inbox.
How It Works
- Submit a FASTA file: One or more soluble protein sequences in standard single-letter code
- Model runs remotely: Sequence-only features are extracted and scored by the ET classifier
- Results by email: Per-residue HS / NS predictions delivered with interactive plots
Available Features
- Sequence-only input
- Extremely randomized trees
- Per-residue HS / NS labels
- FASTA upload
- Email delivery
- Interactive result plots
- Curated HS database
Predict hot spots from sequence alone: upload a FASTA file and receive per-residue predictions & interactive visualisations, no structural data needed, all in your browser: