environment.yml

# Dedicated conda environment for the tcren Python library.
#
# Provides Python, the MMseqs2 binary (used by arda and by Phase-B MHC mapping), and a
# C++ toolchain so tcren's own extensions build on install. The TCR-annotation backend
# arda ships on PyPI as `arda-mapper` (a tcren dependency in pyproject.toml, imports as
# `arda`) and is pulled in automatically by `pip install -e .`.
#
#   conda env create -f environment.yml
#   conda activate tcren
#   pip install -e .
#
name: tcren
channels:
  - conda-forge
  - bioconda
dependencies:
  - python=3.12
  - mmseqs2
  - cxx-compiler
  - cmake
  - pip
  - numpy>=1.26
  - scipy>=1.11
  - biopython>=1.84
  - pip:
      - polars>=1.0
      - typer>=0.12
      - requests>=2.31
      # Build deps for tcren's C++ extensions (the conda cxx-compiler/cmake above do the build).
      # arda-mapper (imports as `arda`) is a normal PyPI dependency in pyproject.toml.
      - scikit-build-core>=0.10
      - pybind11>=2.12

# --- Optional peptide-modelling engines (tcren.refine.engines) -----------------------------------
# The built-in `dope` and `ccd` engines need nothing beyond the above. The physics (`openmm`) and
# loop-modelling (`promod3`) engines are the open replacements for FlexPepDock / MODELLER and are
# heavy native deps, so they are NOT installed by default. Enable them per env when needed:
#   conda install -n tcren -c conda-forge openmm pdbfixer        # openmm engine
#   conda install -n tcren -c bioconda openstructure promod3     # promod3 engine
# The FlexPepDock oracle (tcren.refine.oracle_flexpep) is an external licensed binary — point
# $ROSETTA_BIN at it; it is never bundled.