imspy-predictors 0.5.0__tar.gz

imspy_predictors-0.5.0/PKG-INFO

@@ -0,0 +1,110 @@
+Metadata-Version: 2.4
+Name: imspy-predictors
+Version: 0.5.0
+Summary: ML-based predictors for CCS, retention time, and fragment intensity in mass spectrometry.
+License-Expression: MIT
+Author: theGreatHerrLebert
+Author-email: davidteschner@googlemail.com
+Requires-Python: >=3.11,<3.14
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Provides-Extra: all
+Provides-Extra: cuda
+Provides-Extra: koina
+Provides-Extra: training
+Requires-Dist: datasets (>=2.0.0) ; extra == "all"
+Requires-Dist: datasets (>=2.0.0) ; extra == "training"
+Requires-Dist: imspy-core (>=0.4.0)
+Requires-Dist: koinapy (>=0.0.10) ; extra == "all"
+Requires-Dist: koinapy (>=0.0.10) ; extra == "koina"
+Requires-Dist: numba (>=0.57.0)
+Requires-Dist: numpy (>=1.21.0)
+Requires-Dist: pandas (>=1.3.0)
+Requires-Dist: scikit-learn (>=1.0)
+Requires-Dist: scipy (>=1.7.1)
+Requires-Dist: torch (>=2.0.0)
+Requires-Dist: torch (>=2.0.0) ; extra == "cuda"
+Requires-Dist: tqdm (>=4.66)
+Requires-Dist: wandb (>=0.12.1) ; extra == "all"
+Requires-Dist: wandb (>=0.12.1) ; extra == "training"
+Description-Content-Type: text/markdown
+
+# imspy-predictors
+
+ML-based predictors for CCS, retention time, and fragment intensity in mass spectrometry.
+
+## Installation
+
+```bash
+pip install imspy-predictors
+```
+
+For remote model access via Koina servers:
+
+```bash
+pip install imspy-predictors[koina]
+```
+
+## Features
+
+- **CCS Prediction**: Deep learning models for collision cross section / ion mobility prediction
+- **Retention Time Prediction**: GRU-based retention time predictors
+- **Fragment Intensity Prediction**: Prosit 2023 timsTOF intensity predictor
+- **Charge State Prediction**: Binomial and deep learning charge state distribution models
+- **Koina Integration**: Access remote prediction models via Koina servers (optional)
+
+## Quick Start
+
+```python
+from imspy_predictors import (
+    load_deep_ccs_predictor,
+    load_deep_retention_time_predictor,
+    Prosit2023TimsTofWrapper,
+)
+
+# Load CCS predictor
+ccs_model = load_deep_ccs_predictor()
+
+# Load RT predictor
+rt_model = load_deep_retention_time_predictor()
+
+# Load intensity predictor
+intensity_model = Prosit2023TimsTofWrapper()
+```
+
+## Submodules
+
+- **ccs/**: CCS / ion mobility prediction
+- **rt/**: Retention time prediction
+- **intensity/**: Fragment intensity prediction (Prosit)
+- **ionization/**: Charge state distribution prediction
+- **koina_models/**: Koina remote model access (requires `koinapy`)
+- **utilities/**: Tokenizers for ML models
+
+## Dependencies
+
+- **imspy-core**: Core data structures (required)
+- **TensorFlow**: Deep learning framework (required)
+- **dlomix**: Deep learning for omics (required)
+- **koinapy**: Koina API client (optional, for remote models)
+
+## Optional Dependencies
+
+Some functionality requires additional packages:
+
+- **imspy-search**: For PSM-based predictions using sagepy
+- **imspy-simulation**: For simulation utilities (e.g., flatten_prosit_array)
+
+## Related Packages
+
+- **imspy-core**: Core data structures and timsTOF readers
+- **imspy-search**: Database search functionality
+- **imspy-simulation**: Simulation tools for timsTOF data
+- **imspy-vis**: Visualization tools
+
+## License
+
+MIT License - see LICENSE file for details.
+

imspy_predictors-0.5.0/README.md

@@ -0,0 +1,76 @@
+# imspy-predictors
+
+ML-based predictors for CCS, retention time, and fragment intensity in mass spectrometry.
+
+## Installation
+
+```bash
+pip install imspy-predictors
+```
+
+For remote model access via Koina servers:
+
+```bash
+pip install imspy-predictors[koina]
+```
+
+## Features
+
+- **CCS Prediction**: Deep learning models for collision cross section / ion mobility prediction
+- **Retention Time Prediction**: GRU-based retention time predictors
+- **Fragment Intensity Prediction**: Prosit 2023 timsTOF intensity predictor
+- **Charge State Prediction**: Binomial and deep learning charge state distribution models
+- **Koina Integration**: Access remote prediction models via Koina servers (optional)
+
+## Quick Start
+
+```python
+from imspy_predictors import (
+    load_deep_ccs_predictor,
+    load_deep_retention_time_predictor,
+    Prosit2023TimsTofWrapper,
+)
+
+# Load CCS predictor
+ccs_model = load_deep_ccs_predictor()
+
+# Load RT predictor
+rt_model = load_deep_retention_time_predictor()
+
+# Load intensity predictor
+intensity_model = Prosit2023TimsTofWrapper()
+```
+
+## Submodules
+
+- **ccs/**: CCS / ion mobility prediction
+- **rt/**: Retention time prediction
+- **intensity/**: Fragment intensity prediction (Prosit)
+- **ionization/**: Charge state distribution prediction
+- **koina_models/**: Koina remote model access (requires `koinapy`)
+- **utilities/**: Tokenizers for ML models
+
+## Dependencies
+
+- **imspy-core**: Core data structures (required)
+- **TensorFlow**: Deep learning framework (required)
+- **dlomix**: Deep learning for omics (required)
+- **koinapy**: Koina API client (optional, for remote models)
+
+## Optional Dependencies
+
+Some functionality requires additional packages:
+
+- **imspy-search**: For PSM-based predictions using sagepy
+- **imspy-simulation**: For simulation utilities (e.g., flatten_prosit_array)
+
+## Related Packages
+
+- **imspy-core**: Core data structures and timsTOF readers
+- **imspy-search**: Database search functionality
+- **imspy-simulation**: Simulation tools for timsTOF data
+- **imspy-vis**: Visualization tools
+
+## License
+
+MIT License - see LICENSE file for details.

imspy_predictors-0.5.0/pyproject.toml

@@ -0,0 +1,55 @@
+[project]
+name = "imspy-predictors"
+version = "0.5.0"
+description = "ML-based predictors for CCS, retention time, and fragment intensity in mass spectrometry."
+authors = [
+    { name = "theGreatHerrLebert", email = "davidteschner@googlemail.com" }
+]
+readme = "README.md"
+license = "MIT"
+requires-python = ">=3.11,<3.14"
+
+dependencies = [
+    "imspy-core>=0.4.0",
+    # PyTorch - deep learning framework
+    "torch>=2.0.0",
+    # Scientific computing
+    "scikit-learn>=1.0",
+    "scipy>=1.7.1",
+    "numpy>=1.21.0",
+    "pandas>=1.3.0",
+    # Utilities
+    "tqdm>=4.66",
+    "numba>=0.57.0",
+]
+
+[project.optional-dependencies]
+# Koina server access for remote predictions (including intensity via Prosit)
+koina = ["koinapy>=0.0.10"]
+# Hugging Face datasets for training
+training = ["datasets>=2.0.0", "wandb>=0.12.1"]
+# GPU support
+cuda = ["torch>=2.0.0"]
+# All optional dependencies
+all = [
+    "koinapy>=0.0.10",
+    "datasets>=2.0.0",
+    "wandb>=0.12.1",
+]
+
+[tool.poetry]
+exclude = ["src/imspy_predictors/pretrained/**/*.pt"]
+
+[build-system]
+requires = ["poetry-core"]
+build-backend = "poetry.core.masonry.api"
+
+[tool.poetry.group.dev.dependencies]
+pytest = "^8.0.0"
+pytest-cov = "^4.1.0"
+
+[tool.pytest.ini_options]
+testpaths = ["tests"]
+python_files = ["test_*.py"]
+python_functions = ["test_*"]
+addopts = "-v --tb=short"

imspy_predictors-0.5.0/src/imspy_predictors/__init__.py

@@ -0,0 +1,280 @@
+"""
+imspy_predictors - ML-based predictors for CCS, retention time, and fragment intensity in mass spectrometry.
+
+This package provides machine learning models for predicting peptide properties:
+- CCS (Collision Cross Section) / Ion Mobility
+- Retention Time
+- Fragment Intensity (via Koina/Prosit)
+- Charge State / Ionization
+
+All models use PyTorch as the deep learning backend.
+
+Requires imspy-core for core data structures and Rust tokenizer.
+
+Optional dependencies:
+- koina: For remote model access via Koina servers (pip install imspy-predictors[koina])
+- imspy-search: For sagepy-based PSM predictions
+- imspy-simulation: For simulation utilities
+"""
+
+__version__ = "0.5.0"
+
+# Track which components are available
+_IMSPY_CORE_AVAILABLE = False
+_TORCH_AVAILABLE = False
+
+# Core utility functions (no external deps)
+from imspy_predictors.utility import (
+    get_model_path,
+    load_tokenizer_from_resources,
+    InMemoryCheckpoint,
+    get_device,
+    count_parameters,
+    save_model_checkpoint,
+    load_model_checkpoint,
+)
+
+# Hashing utilities (PyTorch only, no imspy_core)
+from imspy_predictors.hashing import (
+    CosimHasher,
+    TimsHasher,
+    SpectralHasher,
+)
+
+# Mixture models (PyTorch only, no imspy_core)
+from imspy_predictors.mixture import (
+    GaussianMixtureModel,
+)
+
+# CCS / Ion Mobility predictors (requires imspy_core)
+try:
+    from imspy_predictors.ccs import (
+        PeptideIonMobilityApex,
+        DeepPeptideIonMobilityApex,
+        SquareRootProjectionLayer,
+        load_deep_ccs_predictor,
+        get_sqrt_slopes_and_intercepts,
+        predict_inverse_ion_mobility_with_koina,
+    )
+    _IMSPY_CORE_AVAILABLE = True
+except ImportError:
+    PeptideIonMobilityApex = None
+    DeepPeptideIonMobilityApex = None
+    SquareRootProjectionLayer = None
+    load_deep_ccs_predictor = None
+    get_sqrt_slopes_and_intercepts = None
+    predict_inverse_ion_mobility_with_koina = None
+
+# Retention time predictors (requires imspy_core)
+try:
+    from imspy_predictors.rt import (
+        PeptideChromatographyApex,
+        DeepChromatographyApex,
+        load_deep_retention_time_predictor,
+        predict_retention_time_with_koina,
+        linear_map,
+    )
+except ImportError:
+    PeptideChromatographyApex = None
+    DeepChromatographyApex = None
+    load_deep_retention_time_predictor = None
+    predict_retention_time_with_koina = None
+    linear_map = None
+
+# Fragment intensity predictors (requires imspy_core)
+try:
+    from imspy_predictors.intensity import (
+        IonIntensityPredictor,
+        Prosit2023TimsTofWrapper,
+        get_collision_energy_calibration_factor,
+        remove_unimod_annotation,
+        predict_fragment_intensities_with_koina,
+        post_process_predicted_fragment_spectra,
+        get_prosit_intensity_flat_labels,
+    )
+except ImportError:
+    IonIntensityPredictor = None
+    Prosit2023TimsTofWrapper = None
+    get_collision_energy_calibration_factor = None
+    remove_unimod_annotation = None
+    predict_fragment_intensities_with_koina = None
+    post_process_predicted_fragment_spectra = None
+    get_prosit_intensity_flat_labels = None
+
+# Charge state / ionization predictors (requires imspy_core)
+try:
+    from imspy_predictors.ionization import (
+        PeptideChargeStateDistribution,
+        BinomialChargeStateDistributionModel,
+        DeepChargeStateDistribution,
+        load_deep_charge_state_predictor,
+        charge_state_distribution_from_sequence_rust,
+        charge_state_distributions_from_sequences_rust,
+        predict_peptide_flyability_with_koina,
+    )
+except ImportError:
+    PeptideChargeStateDistribution = None
+    BinomialChargeStateDistributionModel = None
+    DeepChargeStateDistribution = None
+    load_deep_charge_state_predictor = None
+    charge_state_distribution_from_sequence_rust = None
+    charge_state_distributions_from_sequences_rust = None
+    predict_peptide_flyability_with_koina = None
+
+# Tokenizers (requires imspy_core/Rust bindings)
+try:
+    from imspy_predictors.utilities import ProformaTokenizer
+except ImportError:
+    ProformaTokenizer = None
+
+# HFProformaTokenizer requires transformers (optional)
+try:
+    from imspy_predictors.utilities import HFProformaTokenizer
+except (ImportError, TypeError):
+    HFProformaTokenizer = None
+
+# New PyTorch models (optional - requires torch)
+try:
+    from imspy_predictors.models import (
+        PeptideTransformer,
+        PeptideTransformerConfig,
+        UnifiedPeptideModel,
+        TaskLoss,
+        CCSHead,
+        RTHead,
+        ChargeHead,
+        IntensityHead,
+        INSTRUMENT_TYPES,
+        INSTRUMENT_TO_ID,
+        get_instrument_id,
+    )
+    from imspy_predictors.data_utils import (
+        PeptideDataset,
+        HuggingFaceDatasetWrapper,
+        create_dataloader,
+        collate_peptide_batch,
+        load_ionmob_dataset,
+        load_prospect_rt_dataset,
+        load_prospect_charge_dataset,
+        load_prospect_ms2_dataset,
+        load_timstof_ms2_dataset,
+    )
+    from imspy_predictors.training import (
+        Trainer,
+        TrainingConfig,
+        EarlyStopping,
+        MetricTracker,
+        train_ccs_model,
+        train_rt_model,
+        train_intensity_model,
+    )
+    _TORCH_AVAILABLE = True
+except ImportError:
+    PeptideTransformer = None
+    PeptideTransformerConfig = None
+    UnifiedPeptideModel = None
+    TaskLoss = None
+    CCSHead = None
+    RTHead = None
+    ChargeHead = None
+    IntensityHead = None
+    INSTRUMENT_TYPES = None
+    INSTRUMENT_TO_ID = None
+    get_instrument_id = None
+    PeptideDataset = None
+    HuggingFaceDatasetWrapper = None
+    create_dataloader = None
+    collate_peptide_batch = None
+    load_ionmob_dataset = None
+    load_prospect_rt_dataset = None
+    load_prospect_charge_dataset = None
+    load_prospect_ms2_dataset = None
+    load_timstof_ms2_dataset = None
+    Trainer = None
+    TrainingConfig = None
+    EarlyStopping = None
+    MetricTracker = None
+    train_ccs_model = None
+    train_rt_model = None
+    train_intensity_model = None
+
+__all__ = [
+    # Version
+    '__version__',
+    # Utility
+    'get_model_path',
+    'load_tokenizer_from_resources',
+    'InMemoryCheckpoint',
+    'get_device',
+    'count_parameters',
+    'save_model_checkpoint',
+    'load_model_checkpoint',
+    # Hashing
+    'CosimHasher',
+    'TimsHasher',
+    'SpectralHasher',
+    # Mixture
+    'GaussianMixtureModel',
+    # CCS
+    'PeptideIonMobilityApex',
+    'DeepPeptideIonMobilityApex',
+    'SquareRootProjectionLayer',
+    'load_deep_ccs_predictor',
+    'get_sqrt_slopes_and_intercepts',
+    'predict_inverse_ion_mobility_with_koina',
+    # RT
+    'PeptideChromatographyApex',
+    'DeepChromatographyApex',
+    'load_deep_retention_time_predictor',
+    'predict_retention_time_with_koina',
+    'linear_map',
+    # Intensity
+    'IonIntensityPredictor',
+    'Prosit2023TimsTofWrapper',
+    'get_collision_energy_calibration_factor',
+    'remove_unimod_annotation',
+    'predict_fragment_intensities_with_koina',
+    'post_process_predicted_fragment_spectra',
+    'get_prosit_intensity_flat_labels',
+    # Ionization
+    'PeptideChargeStateDistribution',
+    'BinomialChargeStateDistributionModel',
+    'DeepChargeStateDistribution',
+    'load_deep_charge_state_predictor',
+    'charge_state_distribution_from_sequence_rust',
+    'charge_state_distributions_from_sequences_rust',
+    'predict_peptide_flyability_with_koina',
+    # Tokenizers
+    'ProformaTokenizer',
+    'HFProformaTokenizer',
+    # PyTorch models (new unified architecture)
+    'PeptideTransformer',
+    'PeptideTransformerConfig',
+    'UnifiedPeptideModel',
+    'TaskLoss',
+    'CCSHead',
+    'RTHead',
+    'ChargeHead',
+    'IntensityHead',
+    'INSTRUMENT_TYPES',
+    'INSTRUMENT_TO_ID',
+    'get_instrument_id',
+    # Data utilities
+    'PeptideDataset',
+    'HuggingFaceDatasetWrapper',
+    'create_dataloader',
+    'collate_peptide_batch',
+    'load_ionmob_dataset',
+    'load_prospect_rt_dataset',
+    'load_prospect_charge_dataset',
+    'load_prospect_ms2_dataset',
+    'load_timstof_ms2_dataset',
+    # Training utilities
+    'Trainer',
+    'TrainingConfig',
+    'EarlyStopping',
+    'MetricTracker',
+    'train_ccs_model',
+    'train_rt_model',
+    'train_intensity_model',
+]

imspy_predictors-0.5.0/src/imspy_predictors/ccs/__init__.py

@@ -0,0 +1,32 @@
+"""CCS (Collision Cross Section) prediction module."""
+
+from imspy_predictors.ccs.predictors import (
+    PeptideIonMobilityApex,
+    DeepPeptideIonMobilityApex,
+    SquareRootProjectionLayer,
+    load_deep_ccs_predictor,
+    get_sqrt_slopes_and_intercepts,
+    predict_inverse_ion_mobility_with_koina,
+)
+
+from imspy_predictors.ccs.utility import (
+    load_tokenizer_from_resources as load_ccs_tokenizer,
+    token_list_from_sequence,
+    tokenize_and_pad,
+)
+
+__all__ = [
+    # Predictors
+    'PeptideIonMobilityApex',
+    'DeepPeptideIonMobilityApex',
+    'SquareRootProjectionLayer',
+    # Loaders
+    'load_deep_ccs_predictor',
+    'load_ccs_tokenizer',
+    # Utilities
+    'get_sqrt_slopes_and_intercepts',
+    'token_list_from_sequence',
+    'tokenize_and_pad',
+    # Koina
+    'predict_inverse_ion_mobility_with_koina',
+]