![Author ORCID: We display the ORCID iD icon alongside authors names on our website to acknowledge that the ORCiD has been authenticated when entered by the user. To view the users ORCiD record click the icon. [opens in a new tab]](https://www.cambridge.org/engage/assets/public/coe/logo/orcid.png){kind=logo}
Abstract
Accurate and Interpretable toxicity prediction remains fundamental in computational chemistry and drug discovery. We propose MolGraphormer, a Transformer-GNN hybrid architecture integrating Graph Neural Network message passing with self-attention mechanisms for molecular property prediction. Our model incorporates substructure-aware embeddings via multi-head attention, edge-conditioned message passing, and hierarchical graph aggregation, enabling both local and global molecular reasoning. Evaluated on the Tox21 benchmark dataset, MolGraphormer achieves competitive performance with F1-Score of 0.6697 and AUC-ROC of 0.7806, while maintaining strong recall (0.7787) for identifying toxic compounds. We employ Monte Carlo Dropout and Temperature Scaling for uncertainty quantification, Combined with uncertainty quantification and attention-based interpretability, MolGraphormer offers a practical framework for drug safety assessment and regulatory toxicology.
Supplementary materials
Title
Calibration Curve Graph
Description
Graph showing model calibration of predicted probabilities versus fraction of positives for baseline, temperature scaled, and MC dropout methods.
Actions
Title
Calibration Metrics Comparison
Description
Bar chart comparing Expected Calibration Error (ECE) and Brier scores for baseline, temperature scaled, and MC dropout models.
Actions
Title
Performance Comparison of Graph-Based Models
Description
Comparison of GCN, GAT, GraphSAGE, GIN, and MolGraphormer on molecular prediction tasks using five metrics: Accuracy, Precision, Recall, F1-Score , and AUC-ROC. Demonstrating MolGraphormer's superior balance between precision and recall across benchmarks.
Actions
Supplementary weblinks
Title
Tox21 Dataset
Description
Kaggle - hosted chemical toxicity data used for ML model training and evaluation in computational chemistry
Actions
View 
