janjoswig / AlgoSB2021

Exercise material for the AlgoSB machine learning winter school in Marseille (France) 2021

Algorithms in structural bioinformatics

Machine learning methods to analyze and predict protein structure, dynamics and function

Exercise material for the practical afternoon session on Monday, 8th of November 2021

The content of this exercise aligns with the morning lecture "Learning models of complex dynamics from simulation data" held by Bettina Keller.

Requirements

The exercise will make use of Jupyter notebooks and requires (a recent version of) the following Python packages:

• matplotlib
• numpy
• scipy
• sklearn
• pyemma
• nglview
• cnnclustering

We recommend to use a Python 3.8 based (virtual) environment for this exercise. Using conda maybe the easiest solution here.

If you want to read up on Python virtual environments, you could start with Python Virtual Environments: A Primer.

conda

If you use conda, a ready to use conda virtual environment with all the requirements installed can be created using the provided environment.yml file:

conda env create -f environment.yml

Then activate the new environment:

conda activate AlgoSB

This is equivalent to a manual creation of a fresh environment followed by an installation of the needed packages:

conda create --name AlgoSB python=3.8 -y
conda activate AlgoSB
conda install matplotlib numpy scipy scikit-learn pyemma nglview -c conda-forge

Please note that the cnnclustering package is only available on PyPi:

pip install cnnclustering

or directly from the development repository on GitHub:

git clone https://github.com/janjoswig/CommonNNClustering.git
cd CommonNNClustering
pip install .

pip

If you genrally prefer pip over conda to manage packages, you can instead install all requirements from the provided requirements.txt file:

pip install -r requirements.txt

Please note, that the installation of pyemma via pip can be sometimes problematic, though.

Google Colab

If you like to use the notebook in Colab, consider installing the requirements via condacolab. Open the notebook in Colab and add the following:

!pip install -q condacolab
import condacolab
condacolab.install()

Then install only the still missing dependencies:

!conda install pyemma nglview -c conda-forge
!pip install cnnclustering