The complete set of scripts used to run the chemical space neural networks described in the paper "Harnessing Chemical Space Neural Networks to Systematically Annotate GPCR ligands" 01 -- allows a user to search the graph neighbourhood of any smiles compound and create figures and .csv's reporting on available data. 02 -- preprocesses the input data and calculates the sparse adjacancy matrix (high-RAM requirements). 03 -- creates pytorch geometric .pt files for the CSNN. 04 -- train the GNN and RF/MLP models and replicate results (time intensive, hyperparameter search, multiple replicates to assess stability of training). 05 -- train the R128 model, which predicts bioactivity labels for all 128 receptors in a single forward pass! 06 -- inference code for the R6 model (on DTI mode, one compound, its neighbourhood, and the receptor ESM representation). Using the pre-trained models (quick to evaluate.) Allows the user to specify any input .csv and judge predictions. 07 -- inference code for the R128 Model, predicting 08 -- contains other scripts used to retrieve datasets and plot certain figures