NNP and custom scripts for GrBP5: pretrained model, training data, and structure-search and AFM-tip construction code

NequIP neural network potential (NNP) for GrBP5 peptides on HOPG, together with the custom scripts used in the study. NNP_files/ contains the training configuration ("training_configuration.yaml"), the pre-trained and deployed model weights ("deployed_best_model.pth"), the full training log, and the training dataset ("training_dataset.extxyz"): positions, forces and energies from 1156 DFT single-point calculations (VASP, optB86b-vdW functional). The model was trained using NequIP version 0.6.0. structure_search/ contains the scripts used to generate the peptide assemblies: generate_conformers.py (random backbone-residue rotations producing clash-free conformers) and assemble_peptide_pairs.py (placement of parallel/antiparallel conformer pairs on a graphite substrate). explicit_tip_creation/ contains the scripts used to build the model AFM tip: step1_carve_framework.py (carving a diamond [111] framework into a conical tip) and step2_passivate.py (passivation of dangling bonds with OH, H, and an apex methyl group). See README.txt for requirements and usage; code is released under the MIT license.