Skip to content

RUNTIME_SETUP.md

Latest commit

 

History

History
173 lines (122 loc) · 4.14 KB

RUNTIME_SETUP.md

File metadata and controls

173 lines (122 loc) · 4.14 KB

AutoScan Runtime Environment Setup

Overview

AutoScan requires a Zero-Trust Dependency Environment with strict version pins to ensure reproducibility and correctness. We provide a pre-configured conda environment that users can activate to run AutoScan without manual dependency management.

Prerequisites

  1. Conda/Miniconda installed and on PATH.

  2. AutoScan repository cloned locally.

    git clone https://github.com/vihaankulkarni29/AutoScan.git
cd AutoScan

Quick Start (Recommended)

Step 1: Create the Runtime Environment

Run this command to provision the autoscan-runtime environment with all dependencies pre-pinned:

conda create -y -n autoscan-runtime -c conda-forge \
  python=3.10 \
  openmm==8.0.0 \
  numpy==1.24.3 \
  biopython>=1.81 \
  scipy>=1.10 \
  pandas>=2.0 \
  typer>=0.9.0 \
  mdtraj>=1.9.9 \
  pyyaml>=6.0 \
  meeko>=0.5.0 \
  rdkit>=2023.9.1 \
  gemmi>=0.6.0 \
  pdbfixer>=1.8 \
  openbabel>=3.1.1

Step 2: Download AutoDock Vina Binary

cd AutoScan
python setup_env.py

This downloads Vina 1.2.7+ into tools/vina.exe (Windows) or tools/vina (Linux/Mac).

Step 3: Activate and Verify

conda activate autoscan-runtime
set CONDA_EXE=C:\path\to\conda.exe  # Windows
# export CONDA_EXE=/path/to/conda  # Linux/Mac
python -c "from autoscan.utils.dependency_check import ensure_dependencies; ensure_dependencies(); print('✅ All dependencies OK')"

If you see ✅ All dependencies OK, you're ready to run AutoScan.

Strict Dependency Manifest

Dependency Version Purpose
Python 3.10 Base runtime
openmm 8.0.0 Molecular Dynamics engine (CUDA/CPU)
numpy 1.24.3 Matrix math for RMSD calculations
biopython ≥1.81 PDB parsing and structure manipulation
scipy ≥1.10 Spatial transforms (Kabsch algorithm)
pandas ≥2.0 Dataframe handling for results
typer ≥0.9.0 CLI interface
mdtraj ≥1.9.9 Trajectory analysis and RMSD
pyyaml ≥6.0 YAML config parsing
meeko ≥0.5.0 Ligand preparation
rdkit ≥2023.9.1 Chemistry toolkit
gemmi ≥0.6.0 Crystallographic utils
pdbfixer ≥1.8 Structure repair
openbabel ≥3.1.1 PDB ↔ PDBQT conversion
AutoDock Vina 1.2.7+ Docking engine (binary in tools/)

Running Tests

Once the environment is active, verify the setup with:

conda activate autoscan-runtime

# Structural validation test
pytest tests/test_structural.py -v

# Phase 1 benchmark (10 targets)
python tests/benchmark_phase1.py

# Phase 1.1 fixed benchmark
python tests/benchmark_phase1_fixed.py

Troubleshooting

"Conda not found"

Ensure conda is on PATH:

# Windows
where conda

# Linux/Mac
which conda

If not found, install Miniconda and add it to PATH.

"OpenBabel/PDBFixer not found"

These are installed in the conda env. Ensure you've activated it:

conda activate autoscan-runtime

"Vina binary missing"

Run:

python setup_env.py

"Dependency check fails"

The ensure_dependencies() function performs zero-trust checks. To auto-repair:

from autoscan.utils import build_dependencies
build_dependencies()

Environment Details

  • OS: Windows, Linux, macOS
  • Python: 3.10 (pinned for openmm compatibility)
  • CUDA: Included in openmm build; falls back to CPU if unavailable
  • Package Manager: Conda-forge
  • Installation Size: ~945 MB (including CUDA toolkit)

Publishing Guide for Users

To distribute AutoScan to end-users:

  1. Provide the setup commands above in your documentation/README.
  2. Environment name: autoscan-runtime (standard across users)
  3. Activation: conda activate autoscan-runtime
  4. First run: python setup_env.py then run tests to validate.

Users should never need to manually manage dependencies—zero-trust checks will catch any issues.

Last Updated: February 17, 2026
Status: Production Ready