An aromatic cage, also known as aromatic box or hydrophobic box, is a structural motif lined by two to five closely packed aromatic amino acid residues (Phe, Tyr, Trp, or His) in a highly hydrophobic binding site, often supplemented by a proximal anionic Asp or Glu residue(s). In proteins possessing aromatic cage domains, the specificity of molecular recognition is principally driven by multivalent cation−π interactions formed at the contact interface between a ligand’s cationic center and the aromatic cage residues and, to a lesser extent, by hydrophobic contacts.

All structures were curated from the PDB (https://www.rcsb.org/).

The complexity of a small molecule can be rationalized by evaluating its saturation (shape complexity, Csp3/[Csp2 + Csp3]) or its total content in chiral carbon atoms (stereochemical or structural complexity, Cstereogenic/Ctotal). P. A. Clemons et al. PANS USA 2010, 107, 18787.

The geometric descriptors of aromatic binding sites, namely pocket depth, volume, hydrophobicity, and druggability score, were calculated with DoGSiteScorer, a grid-based program that uses a Difference of Gaussians algorithm for pocket detection and a support vector machine model for druggability assessment. Volkamer, A. et al. J. Chem. Inf. Model. 2010, 50 (11), 2041–2052; Volkamer, A. et al. J. Chem. Inf. Model. 2012, 52 (2), 360–372.

Pocket descriptor evaluation failed in a few cases for shallow, solvent-exposed binding sites (e.g. PDB ID: 6J2P; UniProt ID: SPP1_YEAST). Thus pocket descriptors for these complexes are displayed as 'None'.

Pocket shown as mesh in PSE files are calculated with DoGSiteScorer, a grid-based program that uses a Difference of Gaussians algorithm for pocket detection. Pocket detection failed in a few cases for shallow, solvent-exposed binding sites (e.g. PDB ID: 6J2P; UniProt ID: SPP1_YEAST). Thus pocket in those PSE files are not displayed as mesh.

We calculated the pocket similarity using the APoc program. APoc (Alignment of Pockets), is an efficient program for large-scale structural comparison of protein pockets. We used the default parameters in APoc, which gives p-values and z-scores for each binding site comparison. APoc Ref: Gao, M. & Skolnick, J.; Bioinformatics 29, 597604 (2013).

We have embedded commonly used third-party plugins into the website to increase the functionality, namely: PDBe Molstar (https://github.com/PDBeurope/pdbe-molstar/) and Michelanglo (https://doi.org/10.1093/bioinformatics/btaa104) for the visualization of protein-ligand complexes; the command-line tool Molconverter (Marvin 20.18.0, 2020, ChemAxon, https://chemaxon.com/)for the depiction of 2D ligand structures, ChemDoodle (https://www.chemdoodle.com/) for structure editing, and RDKit (https://www.rdkit.org/)for similarity-based structure search, and BLAST+ (https://ftp.ncbi.nlm.nih.gov/blast/executables/blast+/LATEST/) for sequence similarity.

Prof. Dr. Stefan Günther (stefan.guenther[at]pharmazie.uni-freiburg.de)

Please cite:

Li, J. et al. AroCageDB: A Web-based Resource For Aromatic Cage Binding Sites and Their Intrinsic Ligands (manuscript in preparation).