Related Experiment Video
Updated: Jun 28, 2026
Quantifying the Brain Metastatic Tumor Micro-Environment using an Organ-On-A Chip 3D Model, Machine Learning, and Confocal Tomography
Published on: August 16, 2020
Interpretable graph neural networks for predicting drug activity in triple-negative breast cancer using
Basab Nath1, Prakhar Consul2, Hemant Kumar Gianey3
1School of Computer Science Engineering & Technology, Bennett University, Greater Noida, Uttar Pradesh, India. basabnath@gmail.com.
Abstract:
Triple-negative breast cancer (TNBC) is an oncology indication with urgent need for drug discovery and limited targeted therapies. This work benchmarked several graph neural network (GNN) models for binary drug-sensitivity prediction in TNBC under conditions of scaffold-aware evaluation. A benchmark dataset of 3433 TNBC drug-cell line samples from 373 distinct compounds was compiled from GDSC and labeled using thresholds on IC[Formula: see text] activity measurements. Five GNN architectures (Graph Convolutional Networks, Graph Attention Network v2, Graph Isomorphism Network with edge features, Message Passing Neural Networks, and Transformer-GCN) were trained with Optuna-based hyperparameter optimization and assessed with Bemis-Murcko scaffold-disjoint splitting. Under the scaffold-aware TNBC benchmark, GATv2 achieved the strongest internal performance with an AUROC of 0.701, followed by Transformer-GCN (0.644) and GCN (0.642). External validation was further conducted using FDA-approved anticancer drugs and a phytochemical library to evaluate out-of-distribution generalization across chemically diverse molecular spaces. On the FDA benchmark, GINE achieved the highest AUROC (0.630), while Transformer-GCN and MPNN demonstrated comparatively stronger ranking capability on the phytochemical dataset, achieving AUROCs of 0.704 and 0.685, respectively. The external validation results highlighted the intrinsic difficulty of scaffold-aware molecular extrapolation in low-resource TNBC prediction. Visualization-based explainability analyses with attribution further confirmed the models were paying attention to chemically intuitive motifs, such as aromatic rings, heteroaromatic motifs, and electronegative active substituents. In summary, we show that scaffold-aware benchmarking, external validation, and interpretability analyses lead to more trustworthy and realistic application of AI methods for TNBC drug sensitivity prediction in the context of low-resource drug data.
Insights
This study benchmarks graph neural network models for predicting drug sensitivity in triple-negative breast cancer (TNBC). Scaffold-aware evaluation revealed GATv2 as top performer, highlighting AI
Area of Science:
- Computational Chemistry
- Bioinformatics
- Artificial Intelligence
Background:
- Triple-negative breast cancer (TNBC) presents a significant challenge in oncology due to a lack of targeted therapies.
- Developing effective drug discovery strategies for TNBC is crucial, necessitating advanced computational approaches.
- Existing drug discovery methods often struggle with the limited data available for TNBC.
Purpose of the Study:
- To benchmark various graph neural network (GNN) models for predicting drug sensitivity in TNBC.
- To evaluate model performance using scaffold-aware metrics and external validation datasets.
- To assess the reliability and interpretability of AI-driven drug sensitivity predictions for TNBC.
Main Methods:
- Compiled a benchmark dataset of 3433 TNBC drug-cell line samples from GDSC.
- Trained five GNN architectures (GCN, GATv2, GIN, MPNN, Transformer-GCN) with hyperparameter optimization.
- Assessed models using Bemis-Murcko scaffold-disjoint splitting and external validation with FDA-approved drugs and a phytochemical library.
- Performed visualization-based explainability analyses to understand model decision-making.
Main Results:
- Under scaffold-aware TNBC benchmarking, GATv2 achieved the highest AUROC (0.701), followed by Transformer-GCN (0.644) and GCN (0.642).
- External validation showed GINE performing best on FDA data (AUROC 0.630), while Transformer-GCN and MPNN excelled on phytochemical data (AUROCs 0.704 and 0.685, respectively).
- Explainability analyses confirmed models focused on chemically relevant molecular motifs.
Conclusions:
- Scaffold-aware benchmarking and external validation are essential for realistic AI application in TNBC drug discovery.
- Graph neural networks show promise for predicting drug sensitivity in TNBC, but generalization remains challenging.
- Interpretable AI methods enhance trust and understanding in computational drug discovery for low-resource indications like TNBC.