Structure-based prediction

GL Challis, J Ravel, CA Townsend. Predictive, structure-based model of amino acid recognition by nonribosomal peptide synthetase adenylation domains. Chem Biol 7 211-224, 2000. 

This section presents the structure of the boron hybrides and is arranged in accordance with the relationship defined by Wade s Rules and expressed by Williams and Rudolph. Thus for the boranes containing six or more pairs of skeletal bonding electrons, the relationship between the structures of the closo-, nido-, arachno-and hypho-species is described. In cases where the parent borane does not exist, examples from heteroboranes with the correctly predicted structure based on Williams coordination number pattern recognition theory (CNPR) of borane structures will be described 70. Treated separately will be mono- and diborane species and also species with more than 12 boron atoms. Although there have been several reviews on the structures of the boranes in recent years none have used the current approach89. ... 

Buchwald, R, and Bodor, N. (2001), A simple, predictive, structure-based skin permeability model, / Pharm. Pharmacol., 53(8), 1087-1098. 

Where protein structures are well established, the match with model complexes is good both structurally and spectroscopically, although the protein bridging units are less symmetrical (175). This agreement then provies a basis for predicting structures based on partial data, such as metal-metal distances from EXAFS or characteristic electronic spectra. Model systems for the Fe—0—Fe proteins have been comprehensively reviewed (173-175) and only the macrocyclic systems will be considered here. 

By far the most complex and technically demanding predictive method based on protein sequence data has to do with structure prediction. The importance of being able to adequately and accurately predict structure based on sequence is rooted in the knowledge that, whereas sequence may specify conformation, the same conformation may be specified by multiple sequences. The ideas that structure is conserved to a much greater extent than sequence and that there is a limited number of backbone motifs (Chothia and Lesk, 1986 Chothia, 1992) indicate that similarities between proteins may not necessarily be detected through traditional, sequence-based methods only. Deducing the relationship between sequence and structure is at the root of the protein-folding problem, and current research on the problem has been the focus of several reviews (Bryant and Altschul, 1995 Eisenhaber et al., 1995 Lemer et al., 1995). 

Plotting atomic radii and electronegativities of metals of the periodic table versus group numbers is a way to predict structure based on these rules (Figure 2.4). This is an example for the cooperative character of the structure factors introduced in Section 2.1. 

Several empirical approaches for NMR spectra prediction are based on the availability of large NMR spectral databases. By using special methods for encoding substructures that correspond to particular parts of the NMR spectrum, the correlation of substructures and partial spectra can be modeled. Substructures can be encoded by using the additive model greatly developed by Pretsch [11] and Clerc [12]. The authors represented skeleton structures and substituents by individual codes and calculation rules. A more general additive model was introduced... 

In this second empirical approach, which has also been used for C NMR spectra, predictions are based on tabulated chemical shifts for classes of structures, and corrected with additive contributions from neighboring functional groups or substructures. Several tables have been compiled for different types of protons. Increment rules can be found in nearly any textbook on NMR spectroscopy. 

The reliability of the in silico models will be improved and their scope for predictions will be broader as soon as more reliable experimental data are available. However, there is the paradox of predictivity versus diversity. The greater the chemical diversity in a data set, the more difficult is the establishment of a predictive structure-activity relationship. Otherwise, a model developed based on compounds representing only a small subspace of the chemical space has no predictivity for compounds beyond its boundaries. 

CA Schiffer, JW Caldwell, PA Kollman, RM Stroud. Prediction of homologous protein structures based on conformational searches and energetics. Pi otems 8 30-43, 1990. 

Mathematical predictive modeling based on predictive equations. Analogous chemical structures. Employers would rely on service life values from other chemicals having analogous chemical structure to the contaminant under evaluation for breakthrough. 

In general the relevance of predictions of structure-function relationships based on molecular modeling and structural bioinformatics are threefold. First they can be used to answer the question of which partners (proteins) could interact. Second, predictions generate new hypotheses about binding site, about molecular mechanisms of activation and interaction between two partners, and can lead to new ideas for pharmacological intervention. The third aim is to use the predictions for structure-based drug design. 

Nelfinavir. Using structure-based design in conjunction with predicted oral pharmacokinetics, NFV was identified and found to have potent inhibition of HIV-1 in vitro with an IC50 in the 2nM range (Kaldor et al. 1997). Clinical trials of NFV revealed robust and sustained reductions in HIV-1 RNA with over half of all subjects attaining a persistent 1.6 logxo reduction at 12 months, in conjunction with a mean increase in CD4 cells of 180-200 per mm (Markowitz et al. 1998). In subjects... 

Benigni R, Richard AM. Quantitative structure-based modeling applied to characterization and prediction of chemical toxicity. Methods 1998 14 264-76. 

Richard AM. Structure-based methods for predicting mutagenicity and carcinogenicity are we there yet Mut Res Fund Mol Mech Mut 1998 400 493-507. 

Early experimental spectroscopic investigations on Rg- XY complexes resulted in contradictory information regarding the interactions within them and their preferred geometries. Rovibronic absorption and LIF spectra revealed T-shaped excited- and ground-state configurations, wherein the Rg atom is confined to a plane perpendicular to the X—Y bond [10, 19, 28-30]. While these results were supported by the prediction of T-shaped structures based on pairwise additive Lennard-Jones or Morse atom-atom potentials, they seemed to be at odds with results from microwave spectroscopy experiments that were consistent with linear ground-state geometries [31, 32]. Some attempts were made to justify the contradictory results of the microwave and optical spectroscopic studies, and... 

A number of approaches to predict ionization based on structure have been published (for a review, see [53]) and some of these are commercially available. Predictions tend to be good for structures with already known and measured functional groups. However, predictions can be poor for new innovative structures. Nevertheless, pfCa predictions can still be used to drive a project in the desired direction and the rank order of the compounds is often correct. More recently training algorithms have also become available which use in-house data to improve the predictions. This is obviously the way forward. 

---