Structure Prediction Methods

A variety of computational methods for the prediction of crystal packing have emerged during the last decade. At least three approaches to constructing low energy crystal packings can be discerned  

Construction of low energy clusters of 10-50 molecules, which can be viewed as the nucleus from which the crystal will eventually grow. The center of such a cluster is assumed to be similar to the final crystal structure. Thus, the crystal structure is to be found by simulating the start of the crystallization process. 

Construction of configurations containing 1-10 molecules, related by the desired symmetry elements, which are then subjected to lattice symmetry to form crystals. As in method 1, nonperiodic clusters are generated first, but here, instead of having a relatively large cluster size, translational symmetry is introduced to simulate a bulk environment. 

We mention these approaches mainly to highlight the most characteristic features of different prediction methods. The efficiency, reliability, and general applicability of each method will vary with the particular implementation. 

The crystal structure prediction program (Molecular Packing 

---

Building sequence profiles or Hidden Markov Models to perform more sensitive homology searches. A sequence profile contains information about the variability of every sequence position, improving structure prediction methods (secondary structure prediction). Sequence profile searches have become readily available through the introduction of PsiBLAST [4]... 

Approaches of de novo predictions, which try to calculate how the structural elements are folded into the 3D-stmcture (tertiary structure) of complete proteins are nowadays far away from reliable large-scale applications. On the other, hand this topic is under strong development indicated by recent successful results at the contest for structural prediction methods CASP4. With the fast growing number of experimentally solved 3D-stmctures of protein and new promising approaches like threading tools combined with experimental structural constraints, one can expect more reliable de novo predictions for 3D-protein structures in the future. 

Table 2.6 Some secondary structure predictive methods currently used. Refer to text for further details...

Pazos, F., Olmea, O., and Valencia, A. (1997) A graphical interface for correlated mutations and other protein structure prediction methods. Comput. Appl. Biosci. 13, 319-321. 

D Fischer, C Barret, K Bryson, A Elofsson, A Godzik, D Jones, KJ Karplus, LA Kelley, RM MacCallum, K Pawowski, B Rost, L Rychlewski, MJ Sternberg. CA-FASP-1 critical assessment of fully automated structure prediction methods. Proteins Structure, Function and Genetics (Suppl) 3 209-217, 1999. 

While the primary structure of proteins and nucleic acids can be experimentally determined in a straight-forward manner, their higher-order structures are much more difficult to elucidate. In general, computational methods dealing with primary structure focus on interpretation of the structure-function, as in promoter analysis. By contrast computational methods working on higher-order structure instead focus on the prediction of structural details. Further, most techniques are limited to the prediction of RNA and protein structures—sugar-, fatty-add-, and DNA-structural prediction methods are in their infancy. 

Protein-structure prediction methods are routinely compared and contrasted in a public competition called the Critical Assessment of Techniques for Protein Structure Prediction (CASP). These large-scale events allow for a comparison of the state-of-the-art tools and algorithms on a variety of target sequences. The results of each event are scrutinized. There have been five such events thus far, with CASP5 being the most recent (31). 

A variety of secondary structure prediction methods has been applied to (3-casein. Regions of a helix around residues 24, 94, and 133 and (3 strands near residues 83, 147, and 190 are widely predicted (Creamer etal., 1981 Graham et al., 1984 Holt and Sawyer, 1988a,b). The predicted a helix in the N-terminal phosphopeptide region may only be stable at low pH, causing the increase in apparent helix content at pH 1.5, compared to neutrality (Creamer et al., 1981). 

Analysis of the CD spectrum has yielded values of 14% a helix and 31 % p strand, with a possible increase in helix content observed with increase of temperature (Loucheaux-Lefebvre et al., 1978). In a more recent study (Ono et al., 1987), a lower fraction of a helix was calculated, but the results vary with the method of calculation. Structure prediction methods have also been applied to this protein and have given results that encourage the view that K-casein has a number of stable conformational features. Loucheaux-Lefebvre et al. (1978) applied the Chou and Fasman (1974) method and predicted an a-helical content of 23%, with 31% P strand and 10% p turns. Raap et al. (1983) preferred the method of Lim (1974) to predict a-helix and P-strand content, because the method of Chou and Fasman, as published in 1974, was considered to overpredict these elements (Lenstra, 1977). They also tested their predictions for the structure about the chymosin-sensitive bond using the later boundary analysis method... 

Source Data from Liljas A, Rossmann MG. X-ray studies of protein interactions. Annu Rev Biochem 43 475-505, 1974 Argos P, Schwarz JS, Schwarz J. An assessment of protein secondary structure prediction methods based on amino acid sequence. Biochim Biophys Acta 439 261-273, 1976. 

Numerous insightful reviews detail the use of the rhodopsin crystal structure as a model building template for other class A GPCRs (Archer et al., 2003 Ballesteros et al., 2001 Becker et al., 2003 FUipek et al, 2003). Modeling studies of GPCR structure have been advanced to such an extent that ab initio structure prediction methods using the minimum possible experimental structural information are underway (Vaidehi et al., 2002). 

Discrimination Between Folds. Because of the inherent error in potential functions, secondary structure prediction methods, limited sampling, and so forth, one can anticipate that prediction of a variety of alternative structures (perhaps, by several methods) would be more likely to generate a correctly folded structure than any single prediction. The problem then becomes one of discriminating between the correct structure and alterna-... 

All the methods developed so far try to extract information, directly or indirectly (Lim, 1974), from the ever growing databases of X-ray crystallography resolved protein structures. Unfortunately, the rate at which new structures are added to the structure databases is far from optimal. Chothia (1992) estimated that all proteins, when their structures are known, would fall into about one thousand folding classes, more than half of them yet to be discovered. If so, this means that a great deal of information in the forthcoming structures is not available for the current methods, and therefore we still must rely on the future to see a coherent and realistic increase in the accuracy of secondary structure prediction methods. 

Table t List of secondary structure prediction methods utilized... 

One key element in the performance analysis of secondary structure prediction methods is the proper selection of the accuracy measurement to be employed. Three different types of predictive accuracy measurements were used (Schultz and Schirmer, 1979) ... 

Figure 3. Multidimensional scaling analysis of the dissimilarities between accuracies of different protein secondary structure prediction methods. The method codes can be found in Table I.

Fischer D, Barret C, Bryson K, Elofsson A, Godzik A, Jones D, et al. CAFASP-1 Critical assessment of fully automated structure prediction methods. Proteins 1999 37(Suppl 3) 209-17. 

Fischer D, Rychlewski L, Dunbrack RL, Jr., Ortiz AR, Elofsson A. CAFASP3 The third critical assessment of fully automated structure prediction methods. Proteins 2003 53(Suppl 6) 503-16. 

Structure-prediction Methods for Inorganic Microporous Crystals... 

---