High-performance framework for Ion-Mobility Spectrometry data processing
rustims is an open-source framework for processing raw data from Ion-Mobility Spectrometry (IMS) in proteomics mass spectrometry. Built with Rust for performance and exposed to Python via PyO3, it combines the speed of compiled code with the flexibility of Python workflows.
- DDA & DIA Support — Full pipelines for both Data-Dependent and Data-Independent Acquisition
- TimSim — Generate synthetic PASEF-like datasets for method development and benchmarking
- Deep Learning Integration — CCS and retention time prediction models
- Rust Performance — Core algorithms implemented in Rust for maximum throughput
- Python API — Easy integration with existing scientific Python workflows
# Install what you need — dependencies are pulled in automatically (Python >=3.11)
pip install imspy-search # DDA/DIA database search (includes core + predictors)
pip install imspy-simulation # TimSim synthetic data generation
pip install imspy-dia # DIA-PASEF clustering
pip install imspy-core # Just the base data structures and timsTOF I/O# Run the DDA analysis pipeline
imspy_dda path/to/data.d path/to/proteome.fasta
# See all options
imspy_dda --helpfrom imspy_core.timstof import TimsDatasetDDA
# Load a dataset
ds = TimsDatasetDDA("path/to/data.d")
# Access frames
frame = ds.get_frame(1)
print(f"Frame 1: {len(frame.mz)} peaks")
# Get spectra for a precursor
spectra = ds.get_pasef_fragments(precursor_id=42)The DIA module provides hierarchical feature extraction with adaptive thresholding:
# Run the DIA clustering pipeline
open_tracer path/to/dia_data.d
# Generate a cluster report
imspy-cluster-report path/to/results/from imspy_core.timstof import TimsDatasetDIA
# Load DIA dataset
ds = TimsDatasetDIA("path/to/dia_data.d")
# Extract features with adaptive noise thresholds
# Thresholds automatically adjust to local signal characteristics
clusters = ds.extract_features(
window_group=1,
adaptive_threshold=True, # Uses N × local noise instead of fixed threshold
sigma_multiplier=3.0 # 3× noise floor
)Adaptive thresholding eliminates hard-coded global thresholds that fail across:
- Different regions of the LC gradient
- Varying m/z ranges with different noise floors
- Samples with different intensity scales
Generate realistic PASEF-like datasets for benchmarking and method development:
# Command line (all paths configured in TOML, with optional overrides)
timsim config.toml
timsim config.toml --save-path output.d --reference-path reference.d --fasta-path proteome.fasta
# Or use the GUI
timsim-guiReal DIA-PASEF (top) vs TimSim synthetic data (bottom)
Built-in models for predicting:
- Collisional Cross Section (CCS) from peptide sequence
- Retention Time (RT) prediction
- Fragment Ion Intensities via Prosit (Adams et al.)
from imspy_predictors.ccs.predictors import PyTorchCCSPredictor
predictor = PyTorchCCSPredictor()
ccs = predictor.predict("PEPTIDEK", charge=2)The integration test framework validates simulated datasets against real proteomics analysis tools — ensuring TimSim produces data that behaves like real experiments.
┌─────────────┐ ┌─────────────┐ ┌─────────────┐ ┌─────────────┐
│ Simulate │ ──▶ │ Analyze │ ──▶ │ Validate │ ──▶ │ Report │
│ (timsim) │ │ (DiaNN/FP) │ │ (vs truth) │ │ (HTML/JSON)│
└─────────────┘ └─────────────┘ └─────────────┘ └─────────────┘
Run a simulation, then evaluate it with production search engines:
# Generate a synthetic DIA-PASEF dataset
python -m imspy_simulation.timsim.integration.sim --env env.toml --test IT-DIA-HELA
# Analyze with DiaNN/FragPipe and validate against ground truth
python -m imspy_simulation.timsim.integration.eval --env env.toml --test IT-DIA-HELAThe pipeline produces HTML reports with pass/fail metrics for identification rate, RT correlation, and ion mobility correlation. Available test scenarios include standard DIA/DDA, multi-species quantification (HYE), phosphoproteomics, and immunopeptidomics.
See the full Validation README for setup, configuration, and adding custom tests.
rust: mscore/, rustdf/, rustms/
pyo3: imspy_connector/
python: packages/ (imspy-core, imspy-predictors, imspy-dia, imspy-search, imspy-simulation, imspy-vis)
julia: imsjl_connector/, IMSJL/ (experimental)
| Crate | Description |
|---|---|
| mscore | Core data structures (spectra, frames, peptides) and algorithms |
| rustdf | TDF file I/O, DIA clustering, peak detection |
| rustms | Additional MS utilities (chemistry, proteomics, isotopes) |
| imspy_connector | PyO3 bindings exposing Rust to Python |
| Package | Description |
|---|---|
| imspy-core | Base data structures, timsTOF dataset access |
| imspy-predictors | PyTorch models for CCS, RT, fragment intensities |
| imspy-dia | DIA-PASEF clustering and feature extraction |
| imspy-search | Database search integration (sagepy, mokapot) |
| imspy-simulation | TimSim synthetic data generation |
| imspy-vis | Interactive visualization tools |
The Rust core provides:
- Memory-efficient data structures for large datasets
- Parallel processing with rayon
- Zero-copy data sharing with Python via NumPy arrays
Install only the packages you need — each one pulls in its dependencies automatically:
| Package | Description | Install |
|---|---|---|
| imspy-core | Core data structures and timsTOF I/O | pip install imspy-core |
| imspy-predictors | ML models for CCS, RT, intensity prediction | pip install imspy-predictors |
| imspy-dia | DIA-PASEF feature extraction and clustering | pip install imspy-dia |
| imspy-search | Database search via sagepy + rescoring | pip install imspy-search |
| imspy-simulation | TimSim synthetic data generation | pip install imspy-simulation |
| imspy-vis | Visualization and plotting tools | pip install imspy-vis |
Dependency tree:
imspy-core (base - required by all)
├── imspy-predictors (adds PyTorch ML models)
│ ├── imspy-search (adds sagepy database search)
│ └── imspy-simulation (adds TimSim)
├── imspy-dia (adds DIA clustering pipeline)
└── imspy-vis (adds visualization)
Example: For DDA analysis with search, install imspy-search which pulls in predictors and core automatically:
pip install imspy-search# Clone repository
git clone https://github.com/theGreatHerrLebert/rustims.git
cd rustims
# Build Rust components
cd mscore && cargo build --release
cd ../rustdf && cargo build --release
# Build Python bindings
pip install maturin[patchelf]
cd ../imspy_connector && maturin build --release
pip install target/wheels/*.whl
# Install Python packages (from packages/ directory)
cd ../packages
pip install -e ./imspy-core
pip install -e ./imspy-predictors
pip install -e ./imspy-dia
pip install -e ./imspy-search
pip install -e ./imspy-simulation
pip install -e ./imspy-visPre-built images available for reproducible environments:
- API Reference: imspy | mscore | rustdf
- Examples: Python packages
- sagepy: Database search examples
If you use rustims in your research, please cite:
Teschner, D., Gomez-Zepeda, D., Łącki, M.K., Kemmer, T., Busch, A., Tenzer, S., and Hildebrandt, A. "Rustims: An Open-Source Framework for Rapid Development and Processing of timsTOF Data-Dependent Acquisition Data." Journal of Proteome Research 24(5), 2358-2368 (2025). DOI: 10.1021/acs.jproteome.4c00966
@article{teschner2025rustims,
title={Rustims: An Open-Source Framework for Rapid Development and Processing of timsTOF Data-Dependent Acquisition Data},
author={Teschner, David and Gomez-Zepeda, David and {\L}{\k{a}}cki, Mateusz K. and Kemmer, Thomas and Busch, Anne and Tenzer, Stefan and Hildebrandt, Andreas},
journal={Journal of Proteome Research},
volume={24},
number={5},
pages={2358--2368},
year={2025},
publisher={American Chemical Society},
doi={10.1021/acs.jproteome.4c00966}
}Contributions are welcome! Whether it's bug reports, feature requests, or code contributions:
- Issues: Open an issue for bugs or feature requests
- Pull Requests: Fork the repo, create a branch, and submit a PR
- Discussions: Join the conversation in GitHub Discussions
MIT License - see LICENSE for details.
Built with Rust and Python | Maintained by the rustims team