GASP

Glycosyltransferase Acceptor Specificity Prediction (GASP). Pan-specific prediction of reactivity between any Glycosyltransferase Superfamily 1 (GT1) and chemical acceptor.

The installation instructions here are thorough in order to allow for reproducing any of the work. The main randomforest train and test script is src/randomforest.py.

REQUIREMENTS

• Unix shell with zsh.
• R.
• Java runtime for some chemical feature generation. E.g. install with homebrew on Mac: brew install --cask temurin
• pip.
• (mini)conda for python 3, e.g. miniforge
• Julia for
  • adding new chemicals to an already established E3FP MDS (chemical features).
  • Feature selection (v1.8.5 used)

INSTALL

• Run ./install.sh. It will install python packages to environment "GT" and julia packages to project-local env. Activate python env with conda activate GT and Julia env e.g. with julia --project=@. when running julia from within the project. install.sh also defines alias git root and calls ./install.R which installs R packages.
• Miller, e.g. with homebrew on Mac: brew install miller

RUN

• Chemical features are generated pubchem IDs or SMILESs with src/chemistry/pipeline.sh. See results/5-chemicalFeatures/ for examples of adding acceptors.
• Adding new enzymes: see results/6-unaligned/ and/or results/7-align/
• encode_features.py --help and randomforest.py --help for instructions.

Troubleshoot

• Scripts or python modules unavailable. Make sure they are available in PATH by modifying $PATH and $PYTHONPATH to easily access code. Assuming zsh this can done by running ./PATHS.sh >> ~/.zshrc or by copy-pasting the output of ./PATHS.sh to somewhere in ~/.zshrc.