Abstrakt
Due to its pharmaceutical applications, one of the most prominent machine learning challenges in bioinformatics is the prediction of drug-target interactions. State-of-the-art approaches are based on various techniques, such as matrix factorization, restricted Boltzmann machines, network-based inference and bipartite local models (BLM).
In this paper, we extend BLM by the incorporation of a hubness-aware regression technique coupled with an enhanced representation of drugs and targets in a multi-modal similarity space. Additionally, we propose to build a projection-based ensemble.
Our Advanced Local Drug-Target Interaction Prediction technique (ALADIN) is evaluated on publicly available realworld drug-target interaction datasets. The results show that our approach statistically significantly outperforms BLM-NII, a recent version of BLM, as well as NetLapRLS and WNN-GIP.