pytorch-frustratometer 0.1.0__tar.gz

pytorch_frustratometer-0.1.0/.gitignore

@@ -0,0 +1,70 @@
+# Python
+__pycache__/
+*.py[cod]
+*.pyo
+*.pyd
+*.so
+*.egg
+*.egg-info/
+dist/
+build/
+.eggs/
+wheels/
+
+# Virtual environments
+.venv/
+venv/
+env/
+
+# Distribution / packaging
+.installed.cfg
+pip-wheel-metadata/
+share/python-wheels/
+MANIFEST
+
+# Testing
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+htmlcov/
+.pytest_cache/
+results/
+
+# Type checkers
+.mypy_cache/
+.dmypy.json
+.pyre/
+
+# Editors
+.vscode/
+.idea/
+*.swp
+*.swo
+*~
+.DS_Store
+*.conflict1
+
+# Jupyter
+.ipynb_checkpoints/
+
+# Project-specific
+*.pdb_singleresidue
+*.pdb_mutational
+*.pdb_configurational
+*.pml
+runs/
+wandb/
+
+*.log
+*.lock
+*.pdb
+*.png
+
+
+# Test fixtures
+!tests/data/*.pdb
+!tests/data/frustratometeR/**/*.pdb_singleresidue
+!tests/data/frustratometeR/**/*.pdb_mutational
+!tests/data/frustratometeR/**/*.pdb_configurational

pytorch_frustratometer-0.1.0/AUTHORS.md

@@ -0,0 +1,18 @@
+# Code Contributors
+
+- Eli J. Draizen <edraizen@gmail.com> ([@edraizen](https://github.com/edraizen))
+- Hana Jaafari ([@HanaJaafari](https://github.com/HanaJaafari))
+- Carlos Bueno <cabb99@gmail.com> ([@cabb99](https://github.com/cabb99))
+- Arnau Llaberia <arnaullaberia27@gmail.com> ([@Zackon123-afk](https://github.com/Zackon123-afk))
+- Ezequiel A. Galpern ([@eagalpern](https://github.com/eagalpern))
+- Franco L. Simonetti <francolsimonetti@gmail.com> ([@francosimonetti](https://github.com/francosimonetti))
+
+# Paper Authors
+
+Eli J. Draizen, Hana Jaafari, Carlos Bueno, Arnau Llaberia, Finley Clark,
+Ezequiel A. Galpern, Franco L. Simonetti, Maria I. Freiberger, Peter G. Wolynes,
+Diego U. Ferreiro, R. Gonzalo Parra
+
+---
+
+Contributions welcome — see [CONTRIBUTING.md](CONTRIBUTING.md).

pytorch_frustratometer-0.1.0/CHANGELOG.md

@@ -0,0 +1,48 @@
+# Changelog
+
+All notable changes to this project will be documented here.
+Format follows [Keep a Changelog](https://keepachangelog.com/en/1.0.0/).
+
+---
+
+## [Unreleased]
+
+### Added
+- Full PyTorch implementation of single-residue, mutational, and configurational frustration modes
+- GPU-accelerated batch processing via `FrustratometerDataset` and `DataLoader`
+- Multi-chain PDB support with chain-aware residue lookup
+- AWSEM potential with burial and water-mediated terms
+- `Coefficients` dataclass for unified hyperparameter management
+- CLI (`pytorch_frustratometer`) with YAML config and per-field overrides
+- PyMOL `.pml` script generation for visualising frustration on 3D structure
+- Plotting subpackage: frustration distributions, outlier maps, FrustraPy/FrustraEvo comparisons
+- Subpackage layout: `algorithm/`, `data/`, `features/`, `structure/`, `utils/`, `plots/`
+- Optional `[plots]` extra for matplotlib/seaborn/umap dependencies
+- Parity tests against Frustratometer web-server and frustratometeR outputs: residue density, native energies
+  and frustration indices in every mode, for single-chain structures and a chain break in contact
+- Exact (decoy-free) mutational frustration: the decoy mean and standard deviation of each contact are the weighted
+  moments over all 20 x 20 amino-acid pairs, as single-residue frustration does over the 20 amino acids
+- Plain functions `decoy_aa_frequencies`, `decoy_aa_weights` and `decoy_statistics`
+  (`pytorch_frustratometer.algorithm`)
+
+### Changed
+- The residue density counts same-chain pairs with `|i-j| > burial_min_sep`.
+  The default `burial_min_sep` is `1`, equivalent to the old `>= 2` convention with default `2`
+- `--min-seq-sep` no longer changes the residue density
+- A `tert_frust_cutoff` (`--cutoff`) shorter than the residue density's reach, `max(well_rmax[0], burial_rmax) +
+  5 / eta` (7.5 A by default), raises an error instead of silently dropping residue pairs from the density
+- Mutational frustration is exact, as single-residue frustration. Results carry no sampling noise, and the decoy
+  standard deviation is the population value (divides by N), as in LAMMPS. `--decoys` sets the decoy count of
+  configurational frustration and of `--use-c-rand` runs only
+- `--use-c-rand` mutational statistics use the population standard deviation, as in LAMMPS
+- MJ mutational frustration honors `--decoy-distribution` and is exact
+- Mutational decoy energies and their background water are evaluated in chunks of `DECOY_CHUNK_ELEMENTS` contacts x decoys, which bounds the memory of large structures with the same results
+
+### Fixed
+- The residue density matches the AWSEM LAMMPS implementation and the Frustratometer web server
+- Decoys are drawn only from valid standard residues, as in LAMMPS. An unknown residue (`aa_type` 20) inside `seq_mask` was drawn as alanine, and with `--use-c-rand` mutational decoys were drawn by tensor position, so a residue excluded from the calculation shifted the draws
+- `Frustratometer("singleresidue" | "mutational" | "configurational")` returns `Frustratometer3`, the model the CLI uses, instead of raising a method-resolution-order `TypeError`
+- `SingleResidueFrustratometer`, `MutationalFrustratometer` and `ConfigurationalFrustratometer` can be built,
+  and so their module CLIs run: their bases repeated a mode mixin that `AWSEM` already holds (`TypeError`).
+  Their densities now count different-chain pairs as `Frustratometer3` does, and the single-residue module
+  CLI no longer stops with a `NameError` after writing its mutation table

pytorch_frustratometer-0.1.0/CONTRIBUTING.md

@@ -0,0 +1,35 @@
+# Contributing
+
+Contributions are welcome and appreciated.
+
+## Getting started
+
+```bash
+git clone https://github.com/ParraLab-ESBG/pytorch-frustratometer.git
+cd pytorch-frustratometer
+pip install -e ".[plots]"
+```
+
+## Reporting bugs
+
+Open an issue on GitHub with:
+- Python and package version (`pip show pytorch-frustratometer`)
+- A minimal reproducible example
+- Full traceback
+
+## Submitting changes
+
+1. Fork the repository and create a branch from `main`.
+2. Make your changes. Add or update tests where relevant.
+3. Ensure existing tests pass: `pytest tests/`
+4. Open a pull request with a clear description of what changed and why.
+
+## Code style
+
+- Follow existing conventions in the file you are editing.
+- Keep functions focused; avoid adding abstractions beyond what the change requires.
+- Default to writing no comments — only add one when the *why* is non-obvious.
+
+## License
+
+By contributing you agree that your work will be released under the [CC BY 4.0](LICENSE) licence.