Data and code for article "Graph Neural Network Prediction of Infrared Spectra of Interstellar Polycyclic Aromatic Hydrocarbons"

Data and code for article "Graph Neural Network Prediction of Infrared Spectra of Interstellar Polycyclic Aromatic Hydrocarbons" USAGE INSTRUCTIONS:-------------------------1. MODEL CODE DESCRIPTION-------------------------This code repository contains several graph neural network (GNN) models for predicting infrared spectral data of polycyclic aromatic hydrocarbons (PAHs). All models are trained using different loss functions. Model Files:AFP model with EMD loss: PAH_EMD_AFP.pyGAT model with EMD loss: PAH_EMD_GAT.pyGCN model with EMD loss: PAH_EMD_GCN.pyMFP model with EMD loss: PAH_EMD_MFP.pyMPNN model with EMD loss: PAH_EMD_MPNN.pyAFP model with HD loss: PAH_HD_AFP.pyAFP model with JSD loss: PAH_JSD_AFP.pyAFP model with SIS loss: PAH_SIS_AFP.pyAFP model with TVD loss: PAH_TVD_AFP.pyAFP generalization test with JSD loss: PAH_JSD_TEST_AFP.py Note: The AFP model performed best in our study, with the JSD loss function version selected as the final optimal model. -------------------------2. DATA DESCRIPTION-------------------------Data Sources:Originally from NASA Ames PAH Database versions 3.2 and 4.0, with selections and modifications.Contains SMILES strings of neutral PAH molecules and their corresponding infrared spectral data Data Files:High-frequency spectra from v3.2: 3.2_CH_Cleaner_ALL_High_PAHs_Dataset.pickleLow-frequency spectra from v3.2: 3.2_CH_Cleaner_PAHs_Dataset.pickleHigh-frequency spectra from v4.0: 4.0_CH_Cleaner_50_100_ALL_High_PAHs_Dataset.pickleLow-frequency spectra from v4.0: 4.0_CH_Cleaner_50_100_PAHs_Dataset.pickleComplete data (high + low frequency) from v3.2: 3.2_PAHs_Data.csvComplete data (high + low frequency) from v4.0: 4.0_PAHs_Data.csv Note: Version 4.0 data is used to test the model's generalization capability to larger molecules (containing 50-100 carbon atoms). -------------------------3. EXECUTION INSTRUCTIONS-------------------------Environment Setup:Ensure required Python packages are installed (PyTorch, RDKit, DeepChem, etc.)Place all code and data files in the same directory Execution Steps:Run any model file (e.g., PAH_JSD_AFP.py) for training or predictionPrediction results will be automatically saved to the "Fold_Predictions" directoryThe "Best_model" file in this directory contains the trained AFP model using JSD lossWhen running PAH_JSD_TEST_AFP.py, this best model will be used for predictions -------------------------4. LIMITATIONS-------------------------The current model only supports neutral PAH molecules and does not support charged molecules or isotopologuesPrediction uncertainty may increase for molecules that differ significantly from the training datasetThe model performs best for PAHs containing 20-40 carbon atoms; prediction accuracy decreases for larger molecules ---------------------------5. Conda env. configuration---------------------------aiohappyeyeballs 2.6.1 aiohttp 3.12.15 aiosignal 1.4.0 async-timeout 5.0.1 attrs 25.3.0 bzip2 1.0.8ca-certificates 2025.7.14cloudpickle 3.1.1 colorama 0.4.6 contourpy 1.3.0 cycler 0.12.1 deepchem 2.8.1.dev20250723140145 dgl 1.1.2 dgllife 0.3.2 filelock 3.18.0 fonttools 4.59.0 frozenlist 1.7.0 fsspec 2025.7.0 future 1.0.0 git 2.49.0huggingface-hub 0.33.4 hyperopt 0.2.7 importlib-resources 6.5.2 jinja2 3.1.6 kiwisolver 1.4.7 libexpat 2.7.1libffi 3.4.6liblzma 5.8.1libsqlite 3.50.3libzlib 1.3.1matplotlib 3.9.4 mpmath 1.3.0 multidict 6.6.4 networkx 3.2.1 openssl 3.5.1pandas 2.3.1 pillow 11.3.0 pip 25.1.1propcache 0.3.2 psutil 7.0.0 py4j 0.10.9.9 pyparsing 3.2.3 python 3.9.23python-dateutil 2.9.0.post0 pytz 2025.2 pyyaml 6.0.2 rdkit 2025.3.3 regex 2024.11.6 safetensors 0.5.3 setuptools 80.9.0sympy 1.14.0 threadpoolctl 3.6.0 tk 8.6.13tokenizers 0.21.2 torch 2.1.2+cu121 torch-cluster 1.6.3+pt21cu121 torch-geometric 2.6.1 torch-scatter 2.1.2+pt21cu121 torch-sparse 0.6.18+pt21cu121 torch-spline-conv 1.2.2+pt21cu121 torchdata 0.7.1 torchvision 0.16.2+cu121 tqdm 4.67.1 transformers 4.53.3 tzdata 2025.2 ucrt 10.0.22621.0 vc 14.3 vc14_runtime 14.44.35208 wheel 0.45.1 yarl 1.20.1