Using HINT to Model Non-covalent Interactions in Electron Density Maps The structure of molecules invisible to the naked eye has become incredibly important both in better understanding biochemical interaction in life forms and in the design of drugs. Based solely upon the structure of a protein, and prior to any expensive and time-consuming wet lab, it is possible (in certain cases) to determine whether drug candidates have any potential binding affinity for their target. With an ever increasing number of anti-biotic resistant strains of pathogenic bacteria, and the emergence of deadly viruses, the development of effective drugs is more pertinent than ever. Already, structural and medicinal chemists have had success in developing specifically designed targeting drugs. A prime example of this process is the development of HIV1 protease inhibitors However, to build such drugs, one must first have a model into which the drug candidate can be fit. Ever since it was utilized by Rosalind Franklin during the discovery of the structure of DNA, x-ray crystallography has been the primary tool utilized by structural chemists in determining DNA/RNA structure and the tertiary and quaternary structures of proteins. The process of x-ray crystallography consists of crystalizing (precipitating the protein into a crystalline structure) the target protein and then bombarding the structure with x-rays. X-rays are diffracted by crystalline structures and are thus scatterd by the electron clouds orbiting the nucleus of the atoms. By measuring the diffraction of the x-rays from many angles, and comparing it with the expected diffraction from the particular crystal structure formed by the protein, one can construct an electron density map. Since the highest concentrations of electrons will be orbiting nuclei, the electron density map provides a 3 dimensional estimate of the position of each atom. Generating the electron density data is only the first step in determining overall structure. Only after threading the known amino acid sequence, atom by atom, through the high density regions of the map is the 3-dimensional structure complete. Despite the fact that x-ray crystallography provides the (theoretically) most accurate view of these microscopic structures the process is imperfect. Several possible problems arise. First, the protein may not fully crystalize (or may not do so at all, in which case the researcher must use NMR). Second, the highest attainable resolution is ~1 angstroms but typical experimental resolutions can be significantly higher than that (i.e. ~2.5 angstroms), at the less precise resolutions, atom position is often ambiguous. Finally, even with high resolution results, determining the precise position and orientation of a side-chain is a difficult process at the early stages of model building. Like so many other fields, the relatively recent surge in computing power has allowed structural chemists to develop tools that partially alleviate these problems. With regard to x-ray crystallography, the primary benefit of such powerful computers has been the generation of programs that can create 3-dimensional models based on experimental data and perform thousands of calculations each second, effectively modeling, in silico, the interactions between each atom. One such set of programs attempts to fit the protein structure to an electron density map. By coding known “rules of interaction” between atoms into such programs the computer can create a best guess at the proper threading of a protein’s sequence through the electron density map. Thus, even with early, rough results, the computer can begin mapping the structure of the molecule before the conclusion of the experiment. There exists an initiative amongst a consortium of labs to generate enough 3-dimensional structures of proteins that a program could potentially generate the tertiary structure of a protein based solely on its similarity to other proteins. However, both a lack of sufficient models and our incomplete understanding of atomic interactions currently inhibit our ability to correctly predict tertiary structure. In fact, even the programs that attempt to thread a protein through an electronic density field make incorrect assertions about potential structures. A simple example: a program may create an appropriate ring-shaped backbone around an active site, but rather than orienting portions of adjacent amino acids away from each other, it may simply face all the ends of the amino acids into the center of the ring. Any knowledgeable human would instantly detect this error, but as far as the program is concerned, it is a perfectly valid configuration. This is because, while the computer is good at modeling electrostatic interactions between atoms, most software packages do a miserable job at accounting for the obfuscated hydrophobic effect. The hydrophobic effect is, essentially, all non-polar interactions between molecules. However, certain researchers, realizing that current models will always be inaccurate if they fail to account for the hydrophobic effect have taken steps to account for it. The HINT program, developed by Glen Kellogg, calculates a score based on noncovalent interactions between each atom. The algorithm used to develop the scores is *equation forthcoming*. The equation is explained as follows: “ai and aj are hydrophobic atom constants for the two atoms being evaluated, S is the solvent accessible surface are, T is a function that differentiates polar-polar interactions (acid-acid, acid-base or base-base), and R, r are functions of the distance between atoms i and j” (Cashman, 120). The HINT program sums the score for each atom to obtain a score for the entire molecule. After the development of HINT, the next step is to utilize the results of the calculations to make existing models more accurate. One program that currently does a good job of generating three dimensional models of proteins from electron density maps is XTalView. The program is especially useful during the beginning stages of the crystallography, before the repetition of many phases. During this time, electron density is noisy, but XTalView can utilize these approximations to create a fairly accurate representation of the molecule. Also, side-chains are difficult to generate because of this noise, and often are only modelled using placeholders and later filled in after acquiring more comprehensive data. However, in threading the amino acid chain through the map, XTalView orients portion of the molecule in a manner that is immediately obvious as incorrect to a researcher examining the data. This project seeks to extend XTalView to consider HINT scores while generating these rough drafts of the final tertiary structure, creating an accurate model from less accurate data. XTalView is a good candidate to interface with HINT because it is extensible (i.e. it was designed so that programmers could expand its functionality) and it already considers several factors during the fitting of the model to the electron density map. These factors are weighted, and thus one could examine the models both with and without the HINT scores, allowing one to judge how the noncovalent interactions help to refine the tertiary structure.