Iterative refinement process, structure

To achieve the greatest improvements in drug discovery efficiency, empirical data of various kinds must be collected throughout the iterative refinement process. It is desirable to obtain more accurate dissociation constants rather than IC50 or single-point percent-inhibition values. In addition, the 3-dimensional structures of interesting target—inhibitor complexes are determined... 

In addition to corrections tising an appropriate spectral density function, in principle one also needs to consider an ensemble of structures. Bonvin et al. U993) used an ensemble iterative relaxation matrix approach in which the NOE is measured as an ensemble property. A relaxation matrix is built from an ensemble of structures, using averaging of contributions from different structures. The needed order parameters for fast motions were obtained fi um a 50-ps molecular dynamics calculation. The relaxation matrix is then used to refine individual structures. The new structures are used again to reconstruct the relaxation matrix, and a second new set of structures is defined. One repeats the process until the ensemble of structures is converged. The caveat espressed earlier that the accuracy of the result is limited by the accuracy of the spectral density function applies to all calculations of this typ . 

For each assumed nuclear arrangement Kj of the iterative structure refinement process, an AFDF electron density distribution p(r, Kj) can be calculated, replacing the conventional Gaussian density representations. The least square fit process using AFDF densities consists of the following formal steps, carried out iteratively ... 

In the above scheme, step B is expected to be the slowest. However, the increased accuracy of the theoretical electronic density is likely to reduce the number of iterations within the structure refinement process substantially. Whereas the estimated overall computer time requirement of the (JCR-AFDF approach is comparable to that of the standard structure refinement method, some improvements in speed as well as in accuracy are expected. 

We anticipate two advantages of using the more realistic electron densities obtained by the AFDF methods. More reliable theoretical electronic charge densities calculated for each assumed nuclear geometry in the course of the iterative structure refinement process will improve the reliability of comparisons with the experimental di action pattern. In particular, AFDF electron densities are expected to serve as more sensitive and more reliable criteria for accepting or rejecting an assumed structure than the locally spherical or possibly elliptical electron density models used in the conventional approach. We also expect that the more accurate density representations within the QCR-AFDF framework will facilitate a more complete utilization and interpretation of the structural information contained in the observed X-ray diffraction pattern. 

With XRD applied to bulk materials, a detailed structural analysis of atomic positions is rather straightforward and routine for structures that can be quite complex (see chapter B 1.9) direct methods in many cases give good results in a single step, while the resulting atomic positions may be refined by iterative fitting procedures based on simulation of the diffraction process. 

In molecular pharmacology research an indirect proof of a structural model is possible by functional examinations, e.g., by molecular biological experiments. Well-selected site directed mutagenesis and their functional characterization allows confirmation or rejection of a molecular protein model. The process is organized as an iterative procedure, where the biological answer of suggested mutations is used to refine the model. The iteration continues until the model... 

Since the phase angles cannot be measured in X-ray experiments, structure solution usually involves an iterative process, in which starting from a rough estimate of the phases, the structure suggested by the electron density map obtained from Eq. (13-3) and the phase computed by Eq. (13-1) are gradually refined, until the computed structure factor amplitudes from Eq. (13-1) converge to the ones observed experimentally. 

In the MEM/Rietveld analysis, each of the observed structure factors of intrinsically overlapped reflections (for instance, 333 and 511 in a cubic system) can be deduced by the structure model based on a free atom model in the Rietveld refinement. In such a case, the obtained MEM charge density will be partially affected by the free atom model used. In order to reduce such a bias, the observed structure factors should be refined based on the deduced structure factors from the obtained MEM charge density. The detail of the process is described in the review article [9,22-24]. In addition, the phased values of structure factors based on the structure model used in Rietveld analysis are used in the MEM analysis. Thus, the phase refinement is also done for the noncentrosymmetric case as P2, of Sc C82 crystal by the iteration of MEM analysis. The detail of the process is also described elsewhere [25]. All of the charge densities shown in this article are obtained through these procedures. 

An additional major problem with natural membrane systems is that the composition of the one-dimensional unit cell may be unknown because of the presence of proteins of unknown structure and composition. Nevertheless, it is possible to model the electron density profile of a multilamellar system to arrive at a crude model for the distribution of lipids and proteins. The models improve considerably if the lipid composition and lipid stmctures are known. In any case, the basic crystallographic structural method is used The model is refined on the basis of comparisons between the calculated and the experimentally determined intensities through an iterative process. 

While this makes perfect chemical sense, it causes a problem. Whatever the source of the phases used to calculate a new electron density map, some features of the model (or old electron density map) will show up in the new map, because the phases dominate its appearance (Figure 15), as mentioned in Section 9.03.9.3. If our model (and hence the phases calculated from it) is correct, this is not a problem, but since the process of structure refinement that we discuss below is iterative, the correctness of the model must be assessed carefully. In crystallography, what you see is what you put in — also known as model bias. 

Using Monte Carlo, the model is then relaxed using the new potential, resulting in a new calculated g r). The process is then iterated until it reaches convergence. This approach, known as Empirical Potential Structure Refinement (EPSR) has proved very powerful in the study of complex liquids, and of the solvation states of molecules in solution. It is particularly successful when it has as target functions a number (though not necessarily the complete set) of DPDFs from the system in question. 

Based on this work, Boelens designed 4-[4-(l-hydroxy-1-methylethyl)-cyclohexylidenyl]butanal, (27), which, when made, did indeed have excellent olfactive properties. Thus, the activity of compound (27) had been correctly predicted and Boelens s model further substantiated. However, if the prediction had been incorrect it should not, as is quite often the case, be ignored and classified as an exception to the rule . It is important that the reasons for the poor prediction are investigated. Outliers or compounds that exhibit unique biological activity can often provide vital clues about the structural requirements for that biological activity. Thus, the development of a SAR is an iterative process, with the information acquired from new materials being used to refine the model. 

---