Repository for the subject CDD
You will need RDKit, scikit-learn. venv or Conda, Jupyter and Git are also very useful.
sudo apt install git
Download and install from [git](https://git-scm.com/downloads)
venv or Conda
Maybe you will need to install it
python3 -m venv ./venv
Linux
wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86\_64.sh
\# you will need probably to run:
chmod u+x Miniconda3-latest-Linux-x86\_64.sh to run this script
./Miniconda3-latest-Linux-x86\_64.shWindows
Download and install [miniconda](https://docs.conda.io/projects/conda/en/latest/user-guide/install/windows.html)
Apple
Download and install [miniconda](https://docs.conda.io/projects/conda/en/latest/user-guide/install/macos.html)
git clone https://github.com/lich-uct/cdd2024.git
\# create environment with rdkit, preferably updated version
\# cdd is name of the environment, change this for whatever you like, jupyter is for our convenience, otherwise you can use plain python scripts
conda create -n cdd -c conda-forge rdkit scikit-learn pandas seaborn jupyter
conda activate cdd
jupyter notebook # if you want to run it locally
jupyter notebook --ip 0.0.0.0 --port XXXX # ip 0s to see jupyter on the Internet, port XXXX as maybe you don't want to run it on default 8888 portThe project, or 'zapoctovy projekt', should involve roughly the following:
- Picking a set of known biologically active structures (approx. 1000 compounds), load and process it using RDkit
- Performing a basic analysis of the set, descriptors, scaffolds, fingerprints, similarities, etc.
- Comparing the properties of your picked set with other sets
- Trying out some standard cheminformatic methods on the sets
- Preparing set visualizations (at least 2)
- Making a basic classifiers/predictors related to your picked and prepared dataset (at least 2, try different algorithms, descriptors, etc. and discuss (different?) results)
On your part, please try to apply the workflows onto your particular set, and adapt them for your particular cases. Compare your results to the example case (provided mTor, Drugbank and/or ZINC sets), and to those of your colleagues. We can discuss the results directly in your Jupyter notebooks. If you get stuck at any point, or want some additional information, please do not hesitate to ask - this course is for you.
The example jupyter notebooks with the workflows will be uploaded into the /exercises folder of the repository.
Make your own folders in the root of the repository; 'dominule' already have one. Place your own data and notebooks and all other things you need in there, and upload them into this repository at your convenience. You already know git, so this should by no problem... git pull, git commit -m "my message", git push , the usual. Please keep the changes within your own directory, having to merge jupyter notebooks is a pain in the ass.