Determining Zn(II) Binding Affinities of the YiiP–Zinc Transporter and Uno Ferro Single Chain (UFsc) Protein with a Novel Modification of the PKA17 Software.

In this paper, we report results of using molecular modeling to assign specific Zn(II) binding affinities to the known binding sites of the YiiP–zinc transporter. YiiP is a cation-diffusion facilitator. It facilitates the transmembrane exchange of Zn(II) ions and protons. The crystal structure of this protein is known. There are several zinc binding sites, and some of the Zn(II) binding affinities have been measured, but the value of all the binding/dissociation constants and the exact assignment of the sites with these affinities are not completely understood. We have recently developed a fast and accurate coarse-grain framework for predicting protein pKa shifts named PKA17. In this paper, we report extending of the same technique to produce a methodology capable of quickly predicting metal–protein binding affinities. The new software has been named M21. It has been tested on several zinc–protein binding cases, and the average unsigned error in the binding energies has been found to be 2.17 kcal/mol vs. the AMBER average error of 3.49 kcal/mol ( K d ratio of ca. 30 vs. the AMBER one of 330). We have then applied the M21 methodology to calculate and assign the YiiP–Zn(II) binding constants of − 2.31 − 13.28 kcal/mol ( K d values from 2. 0 4 × 1 0 − 2 to 1. 8 5 × 1 0 − 1 0 ). We have also undertaken additional modifications of parameters. On one hand, we have included another 11 zinc binding proteins in our target fitting set. These were the Uno Ferro single chain (UFsc) and its modifications created by the Professor Olga Makhlynets group. On the other hand, we have significantly reduced the number of fittable parameters in order to further reduce the possibility of overfitting and to demonstrate the stability of the technique. The final parameter set has only eight adjustable parameters (as opposed to the above case with 17 independent parameters). The average error for the binding cases compared with the same AMBER test set as above did not change much and was still very low at 2.17 kcal/mol. We believe that these results not only further validate the presented methodology but also point out a promising direction for potential multiple joint experimental and computational collaborative projects. Both PKA17 and M21 software have been deployed with web-based interfaces at http://kaminski.wpi.edu/PKA17/pka%5fcalc.html and http://kaminski.wpi.edu/METAL/metal%5fcalc.html , respectively. We have developed M21, a modification of the previously reported coarse grain PKA17 predictor of pKa values of protein residues. This new M21 software is designed to quickly assess protein-metal ion binding affinities. M21 and PKA17 can be run by using a website interface at http://kaminski.wpi.edu/METAL/metal%5fcalc.html and http://kaminski.wpi.edu/PKA17/pka%5fcalc.html , respectively. [ABSTRACT FROM AUTHOR]