DeepSurf: A surface-based deep learning approach for the prediction of ligand binding sites on proteins

  • 2020-02-13 17:22:39
  • Stelios K. Mylonas, Apostolos Axenopoulos, Petros Daras
  • 1


The knowledge of potentially druggable binding sites on proteins is animportant preliminary step towards the discovery of novel drugs. Thecomputational prediction of such areas can be boosted by following the recentmajor advances in the deep learning field and by exploiting the increasingavailability of proper data. In this paper, a novel computational method forthe prediction of potential binding sites is proposed, called DeepSurf.DeepSurf combines a surface-based representation, where a number of 3Dvoxelized grids are placed on the protein's surface, with state-of-the-art deeplearning architectures. After being trained on the large database of scPDB,DeepSurf demonstrates superior performance on two diverse testing datasets, bysurpassing all its main deep learning-based competitors.


Quick Read (beta)

loading the full paper ...