Steps to implement Struct2Graph in 2025

[bdabykov]

1. Create a Conda environment: conda create --name struct2graph python=3.8.19
2. Install Pytorch: pip install pytorch==1.4.0
3. Install Scikit-learn: pip install scikit-learn==1.3.2
4. Create the list of proteins: if you are going to use your own dataset, then you must create your own TXT file with the list of proteins, in the format:
   UniProt ID PDB ID chain ID, example:
   A0FGR8 4P42 A
   A4UGR9 4F14 B
   A5JGM8 5LR0 A
   *** You must have the PDB files of each protein with their corresponding chains.
5. Run the script parse_entries.py: it will create a folder called input1 that contains:

proteins_X_Y.npy → numerical fingerprints per residue
adjacencies_X_Y.npy → residue adjacency matrices
names_X_Y.npy → PDB IDs
seqs_X_Y.npy → amino acid sequences
fingerprint_dict.pickle → dictionary of fingerprint encodings.

6. Run create_examples.py
7. k-fold-CV.py