How to Set Up a Python Environment for Structural Biology (2026)

Why conda, not pip

Python packages can be installed with either pip (Python’s built-in package manager) or conda (Anaconda’s environment manager). For structural biology, conda is strongly preferred for two reasons.

First, structural biology packages like MDAnalysis have compiled C and C++ extensions that pip installs from source — requiring a working compiler toolchain on your machine, which is not always present and frequently causes cryptic error messages. Conda installs pre-compiled binary packages that work immediately on any supported platform.

Second, conda manages full environment isolation. A dedicated structbio environment has its own Python interpreter and package set, completely separate from any other Python on your machine. This means BioPython’s dependencies can’t conflict with your web scraping project’s dependencies, and you can always recreate the environment from scratch if something breaks.

Installing Miniconda

# Download and run the Miniconda installer (Linux example)
wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh
bash Miniconda3-latest-Linux-x86_64.sh

# Follow prompts — accept license, confirm install location
# When asked "Do you wish to initialize Miniconda3?" → yes

# Restart terminal, then verify
conda --version
# Expected output:
conda 24.x.x

Creating the structbio environment

Once Miniconda is installed, create a dedicated environment for all structural biology work. Using a separate environment means you can always return to a clean working state by recreating it, and you’ll never break your base conda installation by installing conflicting packages.

# Create the environment with Python 3.10
conda create -n structbio python=3.10 -y

# Activate it — do this at the start of every work session
conda activate structbio

# Your prompt changes to show the active environment:
(structbio) $

# Deactivate when done
conda deactivate

Installing the essential packages

With the environment active, install all structural biology packages in a single conda command. Installing everything at once lets conda resolve the full dependency graph correctly — doing it piecemeal can lead to conflicts.

# Install all essential packages in one command
conda install -c conda-forge \
    biopython \
    mdanalysis \
    numpy \
    pandas \
    matplotlib \
    scipy \
    jupyter \
    -y

# This takes 3–10 minutes depending on connection speed
# conda-forge has the most up-to-date builds of all these packages

Setting up Jupyter notebooks

Jupyter notebooks let you run Python code in interactive cells and see plots inline — ideal for exploratory analysis where you want to try different selections or visualization parameters without rerunning the entire script. For automated pipelines that run on servers or process large datasets, plain .py scripts are more appropriate.

# Launch Jupyter — opens in your default browser
jupyter notebook

# Or use JupyterLab (more modern interface)
conda install -c conda-forge jupyterlab -y
jupyter lab

# To run on a remote server (HPC cluster) without a browser:
jupyter notebook --no-browser --port=8888
# Then SSH tunnel from your local machine:
# ssh -L 8888:localhost:8888 username@cluster.university.edu
# Then open http://localhost:8888 in your local browser

When Jupyter opens, create a new notebook with the Python 3 kernel. The kernel should automatically use your structbio environment if you launched Jupyter from within it. If you see “kernel not found” or imports fail, confirm that you activated the environment before running jupyter notebook.

Verifying the installation

Run these checks in a new Python session or Jupyter notebook cell to confirm everything is installed and working before starting your first analysis:

import Bio
import MDAnalysis as mda
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt

print(f"BioPython:    {Bio.__version__}")
print(f"MDAnalysis:   {mda.__version__}")
print(f"NumPy:        {np.__version__}")
print(f"pandas:       {pd.__version__}")

# Quick functional test — parse a structure from the PDB
from Bio.PDB import PDBList, PDBParser

pdbl = PDBList()
pdbl.retrieve_pdb_file("1ubq", file_type="pdb", pdir=".")

parser = PDBParser(QUIET=True)
structure = parser.get_structure("ubq", "./ub/pdb1ubq.ent")
chain = structure[0]["A"]
print(f"Ubiquitin: {len(list(chain.get_residues()))} residues loaded")
# Expected output:
Ubiquitin: 76 residues loaded

Optional: adding PyMOL to the same environment

If you also do PyMOL visualization work, installing it in the same structbio environment means your analysis scripts can call PyMOL commands directly — loading a structure in BioPython, filtering it, then passing it to PyMOL for figure generation in one pipeline:

# Add open-source PyMOL to the same environment
conda install -c conda-forge pymol-open-source -y

# Verify
python -c "from pymol import cmd; print('PyMOL available in structbio')"

With all three — BioPython, MDAnalysis, and PyMOL — in a single environment, you can write scripts that span the full structural biology workflow: download an AlphaFold structure with BioPython, analyze the trajectory with MDAnalysis, and generate a publication figure with PyMOL, all in one Python file.