Cart 2

Post-translational modification prediction and mutation assessment (HuPTM)

This tool provides a unified interface for protein post-translational modification (PTM) analysis, supporting six tasks: Acetylation (K), Methylation (K/R), N-glycosylation, SUMOylation (K), Ubiquitination (K), and Phosphorylation (S/T/Y). It supports both site-level probability prediction and mutation effect assessment.

The method is based on the PTMAtlas human proteome dataset and integrates protein language model representations with structural features: RSA(1) + pLDDT(1) + SS(3) + angles(6) = 11 dimensions. The model outputs residue-level modification probabilities and provides false positive rate (FPR) references. In mutation mode, it compares WT and mutant probabilities (delta_prob) to help rapidly screen potentially functional variants.

Tool Limitations: 1) The current model does not explicitly incorporate subcellular localization, which may lead to biologically implausible predictions in certain cellular compartment contexts. 2) Training data is limited to the human proteome and does not yet support cross-species generalization. Results for non-human proteins should be interpreted cautiously and validated experimentally when possible.

1. Task Type

2. Analysis Mode

3. Protein Sequence Input (up to 10 FASTA entries)

Parsed sequences: 0, total residues: 0

Model Performance Metrics

| Task              |    AUPRC |    AUROC |      MCC |       F1 | Precision |       Sn |       Sp | BestThreshold |
|-------------------|---------:|---------:|---------:|---------:|----------:|---------:|---------:|--------------:|
| acetylation_k     | 0.936642 | 0.932756 | 0.713330 | 0.853308 |  0.871811 | 0.835575 | 0.877139 |         0.465 |
| methylation_k     | 0.934669 | 0.942181 | 0.772595 | 0.857143 |  0.903285 | 0.815486 | 0.941950 |         0.400 |
| methylation_r     | 0.957001 | 0.951187 | 0.777299 | 0.883369 |  0.918070 | 0.851197 | 0.924037 |         0.545 |
| glycosylation_n   | 0.994188 | 0.994670 | 0.943865 | 0.972067 |  0.950243 | 0.994917 | 0.947903 |         0.255 |
| sumoylation_k     | 0.898406 | 0.895389 | 0.628095 | 0.811171 |  0.823311 | 0.799383 | 0.828447 |         0.530 |
| ubiquitination_k  | 0.888687 | 0.886633 | 0.599476 | 0.803348 |  0.788256 | 0.819031 | 0.779989 |         0.470 |
| phosphorylation_st| 0.947765 | 0.946868 | 0.752382 | 0.875490 |  0.880313 | 0.870720 | 0.881618 |         0.515 |
| phosphorylation_y | 0.914053 | 0.916010 | 0.676597 | 0.834328 |  0.853521 | 0.815978 | 0.859964 |         0.425 |

References

Wen B, Wang C, Li K, Han P, Holt MV, Savage SR, Lei JT, Dou Y, Shi Z, Li Y, Zhang B. DeepMVP: deep learning models trained on high-quality data accurately predict PTM sites and variant-induced alterations. Nat Methods. 2025 Sep;22(9):1857-1867. doi: 10.1038/s41592-025-02797-x. Epub 2025 Aug 26. PMID: 40859022; PMCID: PMC12446062.

Last updated: 2026-05-11