Drug-Drug Interaction Classification using Graph-Based and Text-Based Drug Representations Graf ve Metin Tabanli Ilaç Temsilleri Kullanilarak Ilaç-Ilaç Etkilesim Siniflandirmasi
34th Signal Processing and Communications Applications Conference, SIU 2026, İstanbul, Türkiye, 7 - 10 Temmuz 2026, (Tam Metin Bildiri)
- Yayın Türü: Bildiri / Tam Metin Bildiri
- Doi Numarası: 10.1109/siu71813.2026.11636353
- Basıldığı Şehir: İstanbul
- Basıldığı Ülke: Türkiye
- Anahtar Kelimeler: BioSentVec, Drug-drug interactions classification, GraphBERT
- Ankara Üniversitesi Adresli: Evet
Özet
Drug-drug interactions (DDI) are important to detect in clinical practice because they can cause side effects and a decrease in treatment effectiveness. This study presents a DDI classification approach that combines graph and text-based drug representations. GraphBERT is used to obtain graph-based drug representations from drugs in the interaction network, while the textual features of the drugs, namely Description, Mechanism, Indication, and Pharmacodynamics, are represented with the BioSentVec model. These representations are used as input to the model for DDI classification separately or in a combined manner. The results show that the combined use of graph and text-based representations provides the highest performance and reaches an F-score of 91.45%. This result was obtained as 0.64% higher than the text-based model and 0.87% higher than the graph-based model. The findings reveal that both representation types provide higher performance when used together.