[albertlieyingadrian]

Fixes #112.

What's wrong

[tool.poetry.dependencies] lists only python, so pip install ScreenPro2 (the README's install command) installs none of the packages the code imports. import screenpro and screenpro --help then fail with ModuleNotFoundError: No module named 'pandas'.

Fix

Declare every third-party package that screenpro/ imports, with pins taken from environment.yml and from what those pins require:

Package	Constraint	Why
pandas	<2.0	same as environment.yml
matplotlib	<3.7	same as environment.yml
polars	>0.20	same as environment.yml
numpy	<2.0	pandas 1.5 is built against numpy 1.x
pydeseq2	<0.5	0.5 requires pandas>=2.2 (as the issue notes)
pyarrow	<26	pyarrow 26 raises ImportError: pyarrow requires NumPy 2.0 or newer at import
python	>=3.9,<3.12	pandas 1.5.3 has no 3.12 wheels, so installs on 3.12 try to build it from source and fail
scipy, bokeh, anndata, scanpy, biobear, statsmodels, scikit-learn, simple_colors, adjustText	unpinned	imported by the package

I left out packages in environment.yml that nothing in screenpro/ imports (seaborn, numba, igraph/leidenalg, watermark, pytest).

Verification

• Clean Python 3.11 venv with pip install . from this branch: resolves to pandas 1.5.3, numpy 1.26.4, matplotlib 3.6.3, pydeseq2 0.4.12, pyarrow 25.0.1, scanpy 1.11.0, anndata 0.11.4, polars 2.1.0. import screenpro works, screenpro --help prints usage, and pytest tests passes 4/4.
• Without the pyarrow<26 cap, pyarrow 26 gets installed and the two test_fastq2count tests fail with the NumPy 2 ImportError above.
• Resolution also works for Python 3.9 and 3.10.
• The built wheel's METADATA now has a Requires-Dist line for each package and Requires-Python: >=3.9,<3.12.

environment.yml still allows python<=3.12 and leaves pydeseq2 unpinned (points 1 and 2 in the issue). I kept this PR to pyproject.toml, but I'm happy to align environment.yml here or in a follow-up if you'd like.

@abearab, could you take a look when you have a moment?

[gemini-code-assist]

Code Review

This pull request updates pyproject.toml to restrict the Python version to >=3.9,<3.12 and explicitly defines several package dependencies. The review feedback identifies a likely typo in the pyarrow version constraint (suggesting <16 instead of <26 to prevent compatibility issues with NumPy 2.0) and recommends replacing wildcard constraints (*) with minimum version requirements to ensure stability.