Combination therapy is widely used in clinical practice, rendering accurate prediction of drug-drug interactions (DDIs) essential for treatment safety and efficacy. This paper proposes PTET-DDI, a novel dual-channel framework that synergizes chemical semantics with geometric structural insights for DDI prediction. Unlike existing approaches that rely solely on 1D sequences or 2D graphs, PTET-DDI integrates a pretrained molecular language model (ChemBERTa) to capture rich context-aware semantic representations, while simultaneously employing an improved fully equivariant graph Transformer to explicitly encode 3D molecular conformations and geometric symmetries. By fusing these complementary modalities, the model not only achieves a comprehensive understanding of molecular properties but also provides interpretability by identifying key structures driving the interactions. Extensive evaluations via 5-fold cross-validation across three benchmark data sets demonstrate that PTET-DDI significantly outperforms existing deep learning-based methods and shows strong generalization capability.
Man et al. (Wed,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: