README.md

---
license: cc-by-nc-sa-4.0
library_name: pytorch
tags:
- proteomics
- mass-spectrometry
- peptide-sequencing
- de-novo-sequencing
- transformer
- biology
- computational-biology
pipeline_tag: text-generation
datasets:
- InstaDeepAI/ms_ninespecies_benchmark
- InstaDeepAI/ms_proteometools
---

# InstaNovo: De novo Peptide Sequencing Model
## Model Description

InstaNovo is a state-of-the-art transformer-based model for de novo peptide sequencing from mass spectrometry data. This model enables accurate, database-free peptide identification for large-scale proteomics experiments. InstaNovo uses a transformer architecture specifically designed for peptide sequencing from tandem mass spectrometry (MS/MS) data. The model predicts peptide sequences directly from MS/MS spectra without requiring a protein database, making it particularly valuable for discovering novel peptides, post-translational modifications, and sequences from organisms with incomplete genomic databases.

## Usage

```python
import torch
import numpy as np
import pandas as pd
from instanovo.transformer.model import InstaNovo
from instanovo.utils import SpectrumDataFrame
from instanovo.transformer.dataset import SpectrumDataset, collate_batch
from torch.utils.data import DataLoader
from instanovo.inference import ScoredSequence
from instanovo.inference import BeamSearchDecoder
from instanovo.utils.metrics import Metrics
from tqdm.notebook import tqdm

# Load the model from the Hugging Face Hub
model, config = InstaNovo.from_pretrained("InstaDeepAI/instanovo-v1.1.0")

# Move the model to the GPU if available
device = "cuda" if torch.cuda.is_available() else "cpu"
model = model.to(device).eval()

# Update the residue set with custom modifications
model.residue_set.update_remapping(
    {
        "M(ox)": "M[UNIMOD:35]",
        "M(+15.99)": "M[UNIMOD:35]",
        "S(p)": "S[UNIMOD:21]",  # Phosphorylation
        "T(p)": "T[UNIMOD:21]",
        "Y(p)": "Y[UNIMOD:21]",
        "S(+79.97)": "S[UNIMOD:21]",
        "T(+79.97)": "T[UNIMOD:21]",
        "Y(+79.97)": "Y[UNIMOD:21]",
        "Q(+0.98)": "Q[UNIMOD:7]",  # Deamidation
        "N(+0.98)": "N[UNIMOD:7]",
        "Q(+.98)": "Q[UNIMOD:7]",
        "N(+.98)": "N[UNIMOD:7]",
        "C(+57.02)": "C[UNIMOD:4]",  # Carboxyamidomethylation
        "(+42.01)": "[UNIMOD:1]",  # Acetylation
        "(+43.01)": "[UNIMOD:5]",  # Carbamylation
        "(-17.03)": "[UNIMOD:385]",
    }
)

# Load the test data
sdf = SpectrumDataFrame.from_huggingface(
    "InstaDeepAI/ms_ninespecies_benchmark",
    is_annotated=True,
    shuffle=False,
    split="test[:10%]",  # Let's only use a subset of the test data for faster inference
)

# Create the dataset
ds = SpectrumDataset(
    sdf,
    model.residue_set,
    config.get("n_peaks", 200),
    return_str=True,
    annotated=True,
)

# Create the data loader
dl = DataLoader(ds, batch_size=64, shuffle=False, num_workers=0, collate_fn=collate_batch)

# Create the decoder
decoder = BeamSearchDecoder(model=model)

# Initialize lists to store predictions and targets
preds = []
targs = []
probs = []

# Iterate over the data loader
for _, batch in tqdm(enumerate(dl), total=len(dl)):
    spectra, precursors, _, peptides, _ = batch
    spectra = spectra.to(device)
    precursors = precursors.to(device)

    # Perform inference
    with torch.no_grad():
        p = decoder.decode(
            spectra=spectra,
            precursors=precursors,
            beam_size=config["n_beams"],
            max_length=config["max_length"],
        )


    preds += [x.sequence if isinstance(x, ScoredSequence) else [] for x in p]
    probs += [
        x.sequence_log_probability if isinstance(x, ScoredSequence) else -float("inf") for x in p
    ]
    targs += list(peptides)

# Initialize metrics
metrics = Metrics(model.residue_set, config["isotope_error_range"])


# Compute precision and recall
aa_precision, aa_recall, peptide_recall, peptide_precision = metrics.compute_precision_recall(
    peptides, preds
)

# Compute amino acid error rate and AUC
aa_error_rate = metrics.compute_aa_er(targs, preds)
auc = metrics.calc_auc(targs, preds, np.exp(pd.Series(probs)))

print(f"amino acid error rate:    {aa_error_rate:.5f}")
print(f"amino acid precision:     {aa_precision:.5f}")
print(f"amino acid recall:        {aa_recall:.5f}")
print(f"peptide precision:        {peptide_precision:.5f}")
print(f"peptide recall:           {peptide_recall:.5f}")
print(f"area under the PR curve:  {auc:.5f}")
```

For more explanation, see the [Getting Started notebook](https://github.com/instadeepai/InstaNovo/blob/main/notebooks/getting_started_with_instanovo.ipynb) in the repository.


## Citation

If you use InstaNovo in your research, please cite:

```bibtex
@article{eloff_kalogeropoulos_2025_instanovo,
        title        = {InstaNovo enables diffusion-powered de novo peptide sequencing in large-scale
                        proteomics experiments},
        author       = {Eloff, Kevin and Kalogeropoulos, Konstantinos and Mabona, Amandla and Morell,
                        Oliver and Catzel, Rachel and Rivera-de-Torre, Esperanza and Berg Jespersen,
                        Jakob and Williams, Wesley and van Beljouw, Sam P. B. and Skwark, Marcin J.
                        and Laustsen, Andreas Hougaard and Brouns, Stan J. J. and Ljungars,
                        Anne and Schoof, Erwin M. and Van Goey, Jeroen and auf dem Keller, Ulrich and
                        Beguir, Karim and Lopez Carranza, Nicolas and Jenkins, Timothy P.},
        year         = {2025},
        month        = {Mar},
        day          = {31},
        journal      = {Nature Machine Intelligence},
        doi          = {10.1038/s42256-025-01019-5},
        issn         = {2522-5839},
        url          = {https://doi.org/10.1038/s42256-025-01019-5}
}
```

## Resources

- **Code Repository**: [https://github.com/instadeepai/InstaNovo](https://github.com/instadeepai/InstaNovo)
- **Documentation**: [https://instadeepai.github.io/InstaNovo/](https://instadeepai.github.io/InstaNovo/)
- **Publication**: [https://www.nature.com/articles/s42256-025-01019-5](https://www.nature.com/articles/s42256-025-01019-5)

## License

- **Code**: Licensed under Apache License 2.0
- **Model Checkpoints**: Licensed under Creative Commons Non-Commercial (CC BY-NC-SA 4.0)

## Installation

```bash
pip install instanovo
```

For GPU support, install with CUDA dependencies:
```bash
pip install instanovo[cu126]
```

## Requirements

- Python >= 3.10, < 3.13
- PyTorch >= 1.13.0
- CUDA (optional, for GPU acceleration)

## Support

For questions, issues, or contributions, please visit the [GitHub repository](https://github.com/instadeepai/InstaNovo) or check the [documentation](https://instadeepai.github.io/InstaNovo/).