Inconsistent numpy versions in the current installation via conda

[sskimb]

I have been using MetaXcan (downloaded on 2024.05.06) for several years.
I need to reinstall MetaXcan on a fresh Ubuntu (actually WSL2).
I cloned the package from github (on 2025.06.30) and tried to install using conda as recommended.

The current conda_env.yaml requires numpy 1.26 and scipy 1.12, as well as bgen-reader 4.0.8.

The scipy 1.12 requires numpy<1.29.0,>=1.22.4. numpy 1.26 satisfies this requirement.

While bgen-reader 4.0.8 requires numpy (>=1.17.0) and cbgen (>=1.0.1), the latter requires numpy (>=2.0,<3.0).

Toward the end of installation, conda installs numpy 2.0.2, which is incompatible with scipy 1.12.

Thank you for your consideration in advance.

The following is the error messages:

MetaXcan/software/metax/gwas/GWAS.py:5: UserWarning: A NumPy version >=1.22.4 and <1.29.0 is required for this version of SciPy (detected version 2.0.2)
  import scipy.stats as stats
Traceback (most recent call last):
  File "MetaXcan/software/SPrediXcan.py", line 13, in <module>
    from metax.gwas import Utilities as GWASUtilities
  File "MetaXcan/software/metax/gwas/Utilities.py", line 7, in <module>
    from . import GWAS
  File "MetaXcan/software/metax/gwas/GWAS.py", line 5, in <module>
    import scipy.stats as stats
  File "miniconda3/envs/imlabtools/lib/python3.9/site-packages/scipy/stats/__init__.py", line 605, in <module>
    from ._stats_py import *
  File "miniconda3/envs/imlabtools/lib/python3.9/site-packages/scipy/stats/_stats_py.py", line 37, in <module>
    from scipy.spatial.distance import cdist
  File "miniconda3/envs/imlabtools/lib/python3.9/site-packages/scipy/spatial/__init__.py", line 110, in <module>
    from ._kdtree import *
  File "miniconda3/envs/imlabtools/lib/python3.9/site-packages/scipy/spatial/_kdtree.py", line 4, in <module>
    from ._ckdtree import cKDTree, cKDTreeNode
  File "_ckdtree.pyx", line 1, in init scipy.spatial._ckdtree
ValueError: numpy.dtype size changed, may indicate binary incompatibility. Expected 96 from C header, got 88 from PyObject

[Linus4]

If you're only using vcf files, you can remove the bgen_reader dependency from the pip section, then create the conda environment. Doesn't work the other way, unfortunately.