Protein Ubiquitination Site Prediction

Deep-learning embeddings from ESM2-650M + gradient-boosted XGBoost — trained on 351,847 experimentally verified sites.

Protein Sequence

FASTA or plain text. Min 51 AA. Only lysine (K) residues are scored.

Load example sequence

AI Model

ESM2-650M + XGBoost

ROC-AUC 0.937 F1 0.814 PR-AUC 0.862

Prediction Threshold 0.45

                                0.30 · sensitive
                                0.70 · strict
                            

Model: ESM-2 + XGBoost

Threshold: 0.45

Training sites: 351,847

Paste Sequence

FASTA or plain amino acids

ESM2-650M Encode

1280-dim layer-33 embeddings

XGBoost Score

K-site probability 0 → 1

Results & Export

CSV / JSON / TSV download