** Update 25.02.2014 ** New potentials and version of program are available

You can download new statistical potentials graphs for Zinc, Calcium and
Magnesium ions or download stand-alone version of program (compiled for Windows 7 (32 bit) or compiled for Linux).

## Method description

Our method for predicting ion binding sites and specificities in protein structures is based on the use of the statistical
potentials (SP) for interaction of proteins with metal ions. Statistical potentials are also sometimes called mean force
potentials and knowledge-based potentials, because they are obtained by statistical analysis of many protein structures,
specifically the data on atom-atom distances for various types of protein atoms, peptide chain dihedral angles, and so on.

We have developed a new method for calculation of statistical potentials
[Rahmanov S.V., Makeev.V.J., Atomic hydration potentials using Monte Carlo reference state advance protein solvation modeling.
FEBS Journal, 2006. 273(s1): p. 62.], using a stochastic technique, simulating non-interacting structure elements as random 3D
points in the structure space. The resulting potentials are detailed, continuous, and cover a wide range of contact distances,
including contacts at short distances. Atomic hydration potentials of proteins obtained with the MCRS method greatly improve
prediction of structure-bound water molecules location in proteins, The potentials provide a means to estimate the total
solvation energy for a protein structure, in many cases achieving a successful fold recognition based solely on the solvation
energy estimates [Rakhmanov, S.V. and V.J. Makeev, Atomic hydration potentials using a Monte Carlo Reference State (MCRS) for
protein solvation modeling. BMC Struct Biol, 2007. 7: p. 19].

Figure 1 below gives some examples of statistical potentials (likelihood ratios) for interaction between several types of atoms which occur in protein structures,
and solvent water oxygen atom.

Figure 1. Statistical potentials (normalized likelihood ratios) for interaction between solvent water oxygen atom and (in the order of decreasing height of the first peak)
Ca++ ions, oxygen atom of structure bound water in proteins, water in bulk liquid.

SP values above 1 indicate preferred atom contact distances (atom interaction energy favors contact formation at the corresponding distance), and vice versa. Values of 1.0 mean
that there an estimate of free energy of making a contact between atoms of involved types, at this distance, equals zero.
Preferable atomic contact distances translate into higher frequencies of atom contacts observed at this distance.

Statistical preferences of structure variables such as atom-atom contact distances in the conformation space
are sometime measured by log likelihood ratio. These preferences can also be measured in energy units, when multiplied
by a kT factor, the absolute temperature times Boltzmann constant. The quasi-energy (the statistical preference)
then takes a reverse form of the Boltzmann equilibrium energy distribution:

E(d) = –kT ln P(d),

Here fobs(d) is the observed frequency of contacts between atoms of two considered types at distance d in the data-
base of macromolecular structures.

Contact preferences of certain protein atoms toward divalent cations have very large first peaks.
Apart from the rest stands interaction of cystein sulfur atom (SG_CYS) with zinc ions (see Figure 2 below),
which demonstrate binding strength close to that of covalent binding.
Figure 2. Distribution of normalized likelihood ratio for contacts between sulfur atom of cystein (SG_CYS) with zinc ions.

## Prediction of ion binding sites in proteins using statistical potentials

Once SPs for interaction between protein atoms and ions are obtained, they can be used for prediction of binding sites and specificities of ions in 3D structures of proteins.
One example of such prediction is given below.

Figure 3. Prediction of Zn++ ion binding site in zinc-binding protein 1FRE.
The experimentally determined position of zinc ion in the structure is shown by a small green sphere.
3D map of local peaks (with a step of ~2Å) of probability maxima for ion placement, calculated using statistical potentials,
are colored according to the estimated probability of ion occurrence, redder part of the spectrum
corresponding to a higher probability.