No abstract

No abstract

Studying the interaction between drugs and targets is the key step of drug repositioning. Through machine learning methods, we can provide reliable drug-target pairs for drug-target interaction (DTI) identification for wet-lab experiments and improve its efficiency. Previous methods did not combine node attributes and relationships of drug and target, which limited the performance of those methods. To this end, we propose a prediction method that takes into account both node attributes and topology information and named it GCNDTI. Using a graph neural network, the low-dimensional feature vectors of drugs and targets are obtained. Utilizing the nonlinear graph neural network model, the drugs and targets’s abstract feature are obtained. The experimental results confirm that compared with added methods, our method shows superior achievement in DTI prediction.

We introduce the concepts of higher-dimensional defect functors and n-Auslander-Reiten translations. Then we prove the higher-dimensional Auslander's defect formula for the category of finitely n-copresented comodules over a left semiperfect coalgebra. Based on this formula, we obtain the higher-dimensional Auslander-Reiten formula for comodule categories.

No abstract