Fig. 1

Visualization of the AGL-EAT graph learning approach. First column: binding site of the molecular complex identified by its PDBID: 3up2. Second column: three different kinds of protein-ligand atom-type pairs. Third column: weighted colored subgraph representations of the corresponding atom-pairs. Fourth column: eigenvalues of the subgraph Laplacian and adjacency matrices. Fifth column: various statistics of these eigenvalues. Fifth column: advanced machine learning models like gradient boosting trees to combine and process these statistics for training and making predictions