Prediction techniques protein modeling

The strongest verification for a 3D-protein model comes from the experimental 3D-structure. This is the objective of the Critical Assessment of Techniques for Protein Structure Prediction, CASP ( http //predic tioncenter.org), where the structural models are made in advance of the experimental structure of a particular protein. 

Even though these approaches are powerful methods for determining functional sites on proteins, they are limited if not coupled with some form of structural determination. As Figure 2 illustrates, molecular biology and synthetic peptide/antibody approaches are not only interdependent, they are tied in with structural determination. Structural determination methods can take many forms, from the classic x-ray crystallography and NMR for three-dimensional determination, to two-dimensional methods such as circular dichroism and Fourier Transformed Infrared Spectroscopy, to predictive methods and modeling. A structural analysis is crucial to the interpretation of experimental results obtained from mutational and synthetic peptide/antibody techniques. 

In the absence of crystallographic or NMR data, predictive techniques based on protein primary sequences can be used to elaborate crude 3D models. Such models will suggest that certain amino acid residues are involved in forming the active (receptor) site. The assignment of structural or functional roles to particular residues can be tested by site-directed mutagenesis, and the model can be further refined by consideration of SAR among ligands. 

The mechanical properties of single hydrated dextran microcapsules (< 10 pm in diameter) with an embedded model protein drug have also been measured by the micromanipulation technique, and the information obtained (such as the Young s modulus) was used to derive their average pore size based on a statistical rubber elasticity theory (Ward and Hadley, 1993) and furthermore to predict the protein release rate (Stenekes et al., 2000). 

The REFINER model had been tested in the CASP5 (every second year community-wide Critical Assessment of Techniques for Protein Structure Prediction) experiment and in a benchmark of de novo prediction of protein fragments. Both tests have shown better performance than that achieved by standard molecular modeling tools. During the next round, CASP6 of the protein structure prediction, the REFINER model produced several very good predictions, especially in the new fold category. 

Interestingly, the diffusional behavior of membrane proteins measured experimentally by FRAP, FCS, or single particle tracking in cells is more complex than predicted by this model. This technique is described best for the case of cell surface proteins, as assessed by FRAP. Such measurements indicate that diffusion is typically much slower than one would expect based on membrane viscosity. In cell membranes, typical values of D for transmembrane proteins are approximately 0.05 pm /s or less, which is much slower than observed in artificial membranes composed of purified lipids. In addition, a significant fraction of proteins is often immobile over the timescale of diffusion experiments (4, 5). Furthermore, diffusional mobilities vary among proteins, and sometimes they differ for the same protein expressed in different cell lines (4, 5). Deviations from pure diffusion are more readily apparent when the trajectories... 

This review will present an overview of current molecular mechanics techniques and discuss some of their limitations. We will then look at knowledge-based protein prediction strategies and examine the incorporation of such empirical rules into refinement methodologies for model protein systems. More comprehensive reviews of molecular dynamics simulations, knowledge-based protein modeling, and protein folding simulations are available. - " ... 

Equally, a model with relatively poor predictivity may be useful in enriching a hit-set with compounds of more favourable properties. Scoring functions for the prediction of protein ligand affinities, even with a prediction error of 4-8 kcal mol have been found to be very useful for identifying actives for potential testing from very large virtual libraries that would be impracticable to screen using experimental techniques. ... 

CALB is an exceptionally robust protein which is deactivated only at 50-60°C, and thus also shows increased resistance towards organic solvents. In contrast to many other lipases, the enzyme appears to be rather rigid and does not show a pronounced effect of interfacial activation [430], which makes it an intermediate between an esterase and a lipase. This latter property is probably the reason why its selectivity could be predicted through computer modeling to a fair extent [431], and for the majority of substrates the Kazlauskas rule (Scheme 2.49) can be applied. In line with these properties of CALB, selectivity-enhancement by addition of water-miscible organic cosolvents such as t-butanol or acetone is possible - a technique which is rather common for esterases. All of these properties make CALB the most widely used lipase both in the hydrolysis [432-437] and synthesis of esters (Sect. 3.1.1). 

---