Structure-Based Function Prediction of Functionally Unannotated Structures in the PDB: Prediction of ATP, GTP, Sialic Acid, Retinoic Acid and Heme-bound and -Unbound (Free) Nitric Oxide Protein Binding Sites
Vicente M. Reyes
DOI: https://doi.org/10.48550/arXiv.1505.01143
IF: 6.064
2015-02-28
Biomolecules
Abstract:Due to increased activity in high-throughput structural genomics efforts around the globe, there has been an accumulation of experimental protein 3D structures lacking functional annotation, thus creating a need for structure-based protein function assignment methods. Computational prediction of ligand binding sites (LBS) is a well-established protein function assignment method. Here we apply the specific LBS detection algorithm we recently described (Reyes, V.M. & Sheth, V.N., 2011; Reyes, V.M., 2015a) to some 801 functionally unannotated experimental structures in the Protein Data Bank by screening for the binding sites (BS) of 6 biologically important ligands: GTP in small Ras-type G-proteins, ATP in ser/thr protein kinases, sialic acid (SIA), retinoic acid (REA), and heme-bound and unbound (free) nitric oxide (hNO, fNO). Validation of the algorithm for the GTP- and ATP-binding sites has been previously described in detail (ibid.); here, validation for the BSs of the 4 other ligands shows both good specificity and sensitivity. Of the 801 structures screened, 8 tested positive for GTP binding, 61 for ATP binding, 35 for SIA binding, 132 for REA binding, 33 for hNO binding, and 10 for fNO binding. Using the cutting plane and tangent sphere methods we described previously, (Reyes, V.M., 2015b), we also determined the depth of burial of the LBSs detected above and compared the values with those from the respective training structures, and the degree of similarity between the two values taken as a further validation of the predicted LBSs. Applying this criterion, we were able to narrow down the predicted GTP-binding proteins to 2, the ATP-binding proteins to 13, the SIA-binding proteins to 2, the REA-binding proteins to 14, the hNO-binding proteins to 4, and the fNO-binding proteins to 1. We believe this further criterion increases the confidence level of our LBS predictions.