Search

Finding structural similarities between proteins often helps reveal shared functionality, which otherwise might not be detected by native sequence information alone. Such similarity is usually detected and quantified by protein structure alignment. Determining the optimal alignment between two protein structures, however, remains a hard problem. An alternative approach is to approximate each three-dimensional protein structure using a sequence of motifs derived from a structural alphabet. Using this approach, structure comparison is performed by comparing the corresponding motif sequences or structural sequences. In this article, we measure the performance of such alphabets in the context of the protein structure classification problem. We consider both local and global structural sequences. Each letter of a local structural sequence corresponds to the best matching fragment to the corresponding local segment of the protein structure. The global structural sequence is designed to generate the best possible complete chain that matches the full protein structure. We use an alphabet of 20 letters, corresponding to a library of 20 motifs or protein fragments having four residues. We show that the global structural sequences approximate well the native structures of proteins, with an average coordinate root mean square of 0.69 A over 2225 test proteins. The approximation is best for all alpha-proteins, while relatively poorer for all beta-proteins. We then test the performance of four different sequence representations of proteins (their native sequence, the sequence of their secondary-structure elements, and the local and global structural sequences based on our fragment library) with different classifiers in their ability to classify proteins that belong to five distinct folds of CATH. Without surprise, the primary sequence alone performs poorly as a structure classifier. We show that addition of either secondary-structure information or local information from the structural sequence considerably improves the classification accuracy. The two fragment-based sequences perform better than the secondary-structure sequence but not well enough at this stage to be a viable alternative to more computationally intensive methods based on protein structure alignment.

Side-chain or even backbone adjustments upon docking of different ligands to the same protein structure, a phenomenon known as induced fit, are frequently observed. Sometimes point mutations within the active site influence the ligand binding of proteins. Furthermore, for homology derived protein structures there are often ambiguities in side-chain placement and uncertainties in loop modeling which may be critical for docking applications. Nevertheless, only very few molecular docking approaches have taken into account such variations in protein structures. We present the new software tool FlexE which addresses the problem of protein structure variations during docking calculations. FlexE can dock flexible ligands into an ensemble of protein structures which represents the flexibility, point mutations, or alternative models of a protein. The FlexE approach is based on a united protein description generated from the superimposed structures of the ensemble. For varying parts of the protein, discrete alternative conformations are explicitly taken into account, which can be combinatorially joined to create new valid protein structures.FlexE was evaluated using ten protein structure ensembles containing 105 crystal structures from the PDB and one modeled structure with 60 ligands in total. For 50 ligands (83 %) FlexE finds a placement with an RMSD to the crystal structure below 2.0 A. In all cases our results are of similar quality to the best solution obtained by sequentially docking the ligands into all protein structures (cross docking). In most cases the computing time is significantly lower than the accumulated run times for the single structures. FlexE takes about five and a half minutes on average for placing one ligand into the united protein description on a common workstation. The example of the aldose reductase demonstrates the necessity of considering protein structure variations for docking calculations. We docked three potent inhibitors into four protein structures with substantial conformational changes within the active site. Using only one rigid protein structure for screening would have missed potential inhibitors whereas all inhibitors can be docked taking all protein structures into account.

The information required to generate a protein structure is contained in its amino acid sequence, but how three-dimensional information is mapped onto a linear sequence is still incompletely understood. Multiple structure alignments of similar protein structures have been used to investigate conserved sequence features but contradictory results have been obtained, due, in large part, to the absence of subjective criteria to be used in the construction of sequence profiles and in the quantitative comparison of alignment results. Here, we report a new procedure for multiple structure alignment and use it to construct structure-based sequence profiles for similar proteins. The definition of "similar" is based on the structural alignment procedure and on the protein structural distance (PSD) described in paper I of this series, which offers an objective measure for protein structure relationships. Our approach is tested in two well-studied groups of proteins; serine proteases and Ig-like proteins. It is demonstrated that the quality of a sequence profile generated by a multiple structure alignment is quite sensitive to the PSD used as a threshold for the inclusion of proteins in the alignment. Specifically, if the proteins included in the aligned set are too distant in structure from one another, there will be a dilution of information and patterns that are relevant to a subset of the proteins are likely to be lost. In order to understand better how the same three-dimensional information can be encoded in seemingly unrelated sequences, structure-based sequence profiles are constructed for subsets of proteins belonging to nine superfolds. We identify patterns of relatively conserved residues in each subset of proteins. It is demonstrated that the most conserved residues are generally located in the regions where tertiary interactions occur and that are relatively conserved in structure. Nevertheless, the conservation patterns are relatively weak in all cases studied, indicating that structure-determining factors that do not require a particular sequential arrangement of amino acids, such as secondary structure propensities and hydrophobic interactions, are important in encoding protein fold information. In general, we find that similar structures can fold without having a set of highly conserved residue clusters or a well-conserved sequence profile; indeed, in some cases there is no apparent conservation pattern common to structures with the same fold. Thus, when a group of proteins exhibits a common and well-defined sequence pattern, it is more likely that these sequences have a close evolutionary relationship rather than the similarities having arisen from the structural requirements of a given fold.

The methodology for generating a homology model of the T1 TCR-PbCS-K(d) class I major histocompatibility complex (MHC) class I complex is presented. The resulting model provides a qualitative explanation of the effect of over 50 different mutations in the region of the complementarity determining region (CDR) loops of the T cell receptor (TCR), the peptide and the MHC's alpha(1)/alpha(2) helices. The peptide is modified by an azido benzoic acid photoreactive group, which is part of the epitope recognized by the TCR. The construction of the model makes use of closely related homologs (the A6 TCR-Tax-HLA A2 complex, the 2C TCR, the 14.3.d TCR Vbeta chain, the 1934.4 TCR Valpha chain, and the H-2 K(b)-ovalbumine peptide), ab initio sampling of CDR loops conformations and experimental data to select from the set of possibilities. The model shows a complex arrangement of the CDR3alpha, CDR1beta, CDR2beta and CDR3beta loops that leads to the highly specific recognition of the photoreactive group. The protocol can be applied systematically to a series of related sequences, permitting the analysis at the structural level of the large TCR repertoire specific for a given peptide-MHC complex.

It is generally accepted that many different protein sequences have similar folded structures, and that there is a relatively high probability that a new sequence possesses a previously observed fold. An indirect consequence of this is that protein design should define the sequence space accessible to a given structure, rather than providing a single optimized sequence. We have recently developed a new approach for protein sequence design, which optimizes the complete sequence of a protein based on the knowledge of its backbone structure, its amino acid composition and a physical energy function including van der Waals interactions, electrostatics, and environment free energy. The specificity of the designed sequence for its template backbone is imposed by keeping the amino acid composition fixed. Here, we show that our procedure converges in sequence space, albeit not to the native sequence of the protein. We observe that while polar residues are well conserved in our designed sequences, non-polar amino acids at the surface of a protein are often replaced by polar residues. The designed sequences provide a multiple alignment of sequences that all adopt the same three-dimensional fold. This alignment is used to derive a profile matrix for chicken triose phosphate isomerase, TIM. The matrix is found to recognize significantly the native sequence for TIM, as well as closely related sequences. Possible application of this approach to protein fold recognition is discussed.

A bank of 13,563 loops from three to eight amino acid residues long, representing motifs between two consecutive regular secondary structures, has been derived from protein structures presenting less than 95 % sequence identity. Statistical analyses of occurrences of conformations and residues revealed length-dependent over-representations of particular amino acids (glycine, proline, asparagine, serine, and aspartate) and conformations (alphaL, epsilon, betaPregions of the Ramachandran plot). A position-dependent distribution of these occurrences was observed for N and C-terminal residues, which are correlated to the nature of the flanking regions. Loops of the same length were clustered into statistically meaningful families on the basis of their backbone structures when placed in a common reference frame, independent of the flanks. These clusters present significantly different distributions of sequence, conformations, and endpoint residue Calphadistances. On the basis of the sequence-structure correlation of this clustering, an automatic loop modeling algorithm was developed. Based on the knowledge of its sequence and of its flank backbone structures each query loop is assigned to a family and target loop supports are selected in this family. The support backbones of these target loops are then adjusted on flanking structures by partial exploration of the conformational space. Loop closure is performed by energy minimization for each support and the final model is chosen among connected supports based upon energy criteria. The quality of the prediction is evaluated by the root-mean-square deviation (rmsd) between the final model and the native loops when the whole bank is re-attributed on itself with a Jackknife test. This average rmsd ranges from 1.1 A for three-residue loops to 3.8 A for eight-residue loops. A few poorly predicted loops are inescapable, considering the high level of diversity in loops and the lack of environment data. To overcome such modeling problems, a statistical reliability score was assigned for each prediction. This score is correlated to the quality of the prediction, in terms of rmsd, and thus improves the selection accuracy of the model. The algorithm efficiency was compared to CASP3 target loop predictions. Moreover, when tested on a test loop bank, this algorithm was shown to be robust when the loops are not precisely delimited, therefore proving to be a useful tool in practice for protein modeling.

The diversity in the lengths and the amino acid sequences of the third complementarity determining region of the antibody heavy chain (CDR-H3) has made it difficult to establish a relationship between the sequences and the tertiary structures, in contrast to the other CDRs, which are classified by their canonical structures. Enhanced conformational sampling of two different CDR-H3s was performed by multicanonical molecular dynamics (multicanonical MD) simulation while restricting the base structures, with and without the other surrounding CDR segments. The results showed that the multicanonical MD sampled a much larger conformational space than the conventional MD, independent of the initial conformations of the simulations. When the other CDRs surrounding the CDR-H3 segments were included in the calculations, the predominant conformations at 300 K corresponded to the X-ray crystal structures. When only the single CDR-H3 loops were considered with the restricted base structures, a greater number of different conformations were sampled as putative loops, but only a small number of stable conformations appeared at 300 K. Analyses of the resultant conformations revealed a structural role for the glycine, when it is located at position three residues before the last residue of CDR-H3 (Gly-X-X-last residue), coincident with the statistical tendencies of many antibody crystal structures. This reflects the general consistency between the energetically stable conformations and the empirically observed conformations. The current method is expected to be applicable to the structural modeling and the design of antibodies, especially for the inherently flexible loops.

Modeling by homology is the most accurate computational method for translating an amino acid sequence into a protein structure. Homology modeling can be divided into two sub-problems, placing the polypeptide backbone and adding side-chains. We present a method for rapidly predicting the conformations of protein side-chains, starting from main-chain coordinates alone. The method involves using fewer than ten rotamers per residue from a backbone-dependent rotamer library and a search to remove steric conflicts. The method is initially tested on 299 high resolution crystal structures by rebuilding side-chains onto the experimentally determined backbone structures. A total of 77% of chi1 and 66% of chi(1 + 2) dihedral angles are predicted within 40 degrees of their crystal structure values. We then tested the method on the entire database of known structures in the Protein Data Bank. The predictive accuracy of the algorithm was strongly correlated with the resolution of the structures. In an effort to simulate a realistic homology modeling problem, 9424 homology models were created using three different modeling strategies. For prediction purposes, pairs of structures were identified which shared between 30% and 90% sequence identity. One strategy results in 82% of chi1 and 72% chi(1 + 2) dihedral angles predicted within 40 degrees of the target crystal structure values, suggesting that movements of the backbone associated with this degree of sequence identity are not large enough to disrupt the predictive ability of our method for non-native backbones. These results compared favorably with existing methods over a comprehensive data set.

We describe a completely automated approach to identifying local sequence motifs that transcend protein family boundaries. Cluster analysis is used to identify recurring patterns of variation at single positions and in short segments of contiguous positions in multiple sequence alignments for a non-redundant set of protein families. Parallel experiments on simulated data sets constructed with the overall residue frequencies of proteins but not the inter-residue correlations show that naturally occurring protein sequences are significantly more clustered than the corresponding random sequences for window lengths ranging from one to 13 contiguous positions. The patterns of variation at single positions are not in general surprising: chemically similar amino acids tend to be grouped together. More interesting patterns emerge as the window length increases. The patterns of variation for longer window lengths are in part recognizable patterns of hydrophobic and hydrophilic residues, and in part less obvious combinations. A particularly interesting class of patterns features highly conserved glycine residues. The patterns provide a means to abstract the information contained in multiple sequence alignments and may be useful for comparison of distantly related sequences or sequence families and for protein structure prediction.

The problem of protein structure prediction is formulated here as that of evaluating how well an amino acid sequence fits a hypothetical structure. The simplest and most complicated approaches, secondary structure prediction and all-atom free energy calculations, can be viewed as sequence-structure fitness problems. Here, an approach of intermediate complexity is described, which involves; (1) description of a protein structure in terms of contact interface vectors, with both intra-protein and protein-solvent contacts counted, (2) derivation of sequence preferences for 2 up to 29 contact interface types, (3) generation of numerous hypothetical model structures by placing the input sequence into a large set of known three-dimensional structures in all possible alignments, (4) evaluation of these models by summing the sequence preferences over all structural positions and (5) choice of predicted three-dimensional structure as that with the best sequence-structure fitness. Evolutionary information is incorporated by using position-dependent core weights derived from multiple sequence alignments. A number of tests of the method are performed: (1) evaluation of cyclic shifts of a sequence in its native structure; (2) alignment of a sequence in its native structure, allowing gaps; (3) alignment search with a sequence or sequence fragment in a database of structures; and (4) alignment search with a structure in a database of sequences. The main results are: (1) a native sequence can very well find its native structure among a large number of alternatives, in correct alignment; (2) substructures, such as (βα)n units, can be detected in spite of very low sequence similarity; (3) remote homologues can be detected, with some dependence on the set of parameters used; (4) contact interface parameters are clearly superior to classical secondary structure parameters; (5) a simple interface description in terms of just two states, protein-protein and protein-water contacts, performs surprisingly well; (6) the use of core weights considerably improves accuracy in detection of remote homologues; (7) based on a sequence database search with a myoglobin contact profile, the C-terminal domain of a vital origin of replication binding protein is predicted to have an all-helical fold. The sequence-structure fitness concept is sufficiently general to accommodate a large variety of protein structure prediction methods, including new models of intermediate complexity currently being developed.

A set of “similarity-parameters” was calculated that reflects the influence of the proteinogenic amino acids on the structure of the protein backbone. The parameters were derived from a detailed analysis of the amino acid specific main-chain torsion angle distributions as they are found in proteins (highly resolved protein structures from the Brookhaven Protein Data Bank). The purpose of these parameters is threefold: (1) they should help in estimating the structural effect of an amino acid substitution during the design of new mutants in protein-engineering; (2) in modeling by homology they should mark places in the protein where changes in the folding are expected; and (3) they should form a scoring matrix in protein sequence alignment superior to identity scoring. The usability of the “structure derived correlation matrix (SCM)” for these purposes is assessed and demonstrated for some examples in the paper.

A great challenge in the proteomics and structural genomics era is to discover protein structure and function, including the identification of biological partners. Experimental investigation is costly and time-consuming, making computational methods very attractive for predicting protein function. In this work, we used the existing structural information in the SH3 family to first extract all SH3 structural features important for binding and then used this information to select the right templates to homology model most of the Saccharomyces cerevisiae SH3 domains. Second, we classified, based on ligand orientation with respect to the SH3 domain, all SH3 peptide ligands into 29 conformations, of which 18 correspond to variants of canonical type I and type II conformations and 11 correspond to non-canonical conformations. Available SH3 templates were expanded by chimera construction to cover some sequence variability and loop conformations. Using the 29 ligand conformations and the homology models, we modelled all possible complexes. Using these complexes and in silico mutagenesis scanning, we constructed position-specific ligand binding matrices. Using these matrices, we determined which sequences will be favorable for every SH3 domain and then validated them with available experimental data. Our work also allowed us to identify key residues that determine loop conformation in SH3 domains, which could be used to model human SH3 domains and do target prediction. The success of this methodology opens the way for sequence-based, genome-wide prediction of protein-protein interactions given enough structural coverage.

Computational protein-protein docking methods currently can create models with atomic accuracy for protein complexes provided that the conformational changes upon association are restricted to the side chains. However, it remains very challenging to account for backbone conformational changes during docking, and most current methods inherently keep monomer backbones rigid for algorithmic simplicity and computational efficiency. Here we present a reformulation of the Rosetta docking method that incorporates explicit backbone flexibility in protein-protein docking. The new method is based on a "fold-tree" representation of the molecular system, which seamlessly integrates internal torsional degrees of freedom and rigid-body degrees of freedom. Problems with internal flexible regions ranging from one or more loops or hinge regions to all of one or both partners can be readily treated using appropriately constructed fold trees. The explicit treatment of backbone flexibility improves both sampling in the vicinity of the native docked conformation and the energetic discrimination between near-native and incorrect models.

We use flexible backbone protein design to explore the sequence and structure neighborhoods of naturally occurring proteins. The method samples sequence and structure space in the vicinity of a known sequence and structure by alternately optimizing the sequence for a fixed protein backbone using rotamer based sequence search, and optimizing the backbone for a fixed amino acid sequence using atomic-resolution structure prediction. We find that such a flexible backbone design method better recapitulates protein family sequence variation than sequence optimization on fixed backbones or randomly perturbed backbone ensembles for ten diverse protein structures. For the SH3 domain, the backbone structure variation in the family is also better recapitulated than in randomly perturbed backbones. The potential application of this method as a model of protein family evolution is highlighted by a concerted transition to the amino acid sequence in the structural core of one SH3 domain starting from the backbone coordinates of an homologous structure.

This article is a personal perspective on the developments in the field of protein folding over approximately the last 40 years. In addition to its historical aspects, the article presents a view of the principles of protein folding with particular emphasis on the relationship of these principles to the problem of protein structure prediction. It is argued that despite much that is new, the essential elements of our current understanding of protein folding were anticipated by researchers many years ago. These elements include the recognition of the central importance of the polypeptide backbone as a determinant of protein conformation, hierarchical protein folding, and multiple folding pathways. Important areas of progress include a detailed characterization of the folding pathways of a number of proteins and a fundamental understanding of the physical chemical forces that determine protein stability. Despite these developments, fold prediction algorithms still encounter difficulties in identifying the correct fold for a given sequence. This may be due to the possibility that the free energy differences between at least a few alternate conformations of many proteins are not large. Significant progress in protein structure prediction has been due primarily to the explosive growth of sequence and structural databases. However, further progress is likely to depend in part on the ability to combine information available from databases with principles and algorithms derived from physical chemical studies of protein folding. An approach to the integration of the two areas is outlined with specific reference to the PrISM program that is a fully integrated sequence/structural-analysis/fold-recognition/homology model building software system.

We report a procedure for the description and comparison of protein surfaces, which is based on a three-dimensional (3D) transposition of the profile method for sensitive protein homology sequence searches. Although the principle of the method can be applied to detect similarities to a single protein surface, the possibility of extending this approach to protein families displaying common structural and/or functional properties, makes it a more powerful tool. In analogy to profiles derived from the multiple alignment of protein sequences, we derive a 3D surface profile from a protein structure or from a multiple structure alignment of several proteins. The 3D profile is used to screen the protein structure database, searching for similar protein surfaces. The application of the procedure to SH2 and SH3 binding pockets and to the nucleotide binding pocket associated with the p-loop structural motif is described. The SH2 and SH3 3D profiles can identify all the SH2 and SH3 binding regions present in the test dataset; the p-loop 3D profile is able to recognize all the p-loop-containing proteins present in the test dataset. Analysis of the p-loop 3D profile allowed the identification of a positive charge whose position is conserved in space but not in sequence. The best ranking non-p-loop-containing protein is an ADP-forming succinyl coenzyme A synthetase, whose nucleotide-binding region has not yet been identified.

An approach to loop prediction that starts with a database search is presented and analyzed. To obtain meaningful statistics, 130 loops from 21 proteins were studied. The correlation between the internal conformation of the loop and the conformation of the neighboring stem residues was examined. Distances between C(alpha) and C(beta) of the immediate neighbor residues at each end select template loops as well as more complex (e.g. three residues on either side) matching criteria. To have a high probability that the best possible loop candidate in the database is included in the set, relatively large cutoffs for matching the interatomic distances of the stem residues have to be used in the template loop selection procedure; for loops of length 5, this results in an average of 1000 loops and for loops of length 9, the number is about 1500. The required number increases only slowly with loop length, in contrast to the exponential time increase involved in direct searches of the conformational space. The best loops among the large number of candidates can be determined by ranking them with the standard CHARMM non-bonded energy function (without electrostatics) applied to the backbone and C(beta) atoms. The same representation (backbone plus C(beta)) can be used to optimize the loop orientations relative to the rest of the protein by constrained energy minimization. Target loops that have many non-bonded contacts with the protein yield better results so that analysis of the non-bonded contacts of the selected template loops is useful in determining the expected accuracy of a prediction. The method for loop selection and optimization predicted eight (out of 18) loops of up to nine residues to an RMSD better than 1.07 A relative to the crystal structure; for 17 of the 18 loops, one of the three lowest energy template loops had an RMSD of less than 1.79 A. The prediction of antibody loops from a database search is more effective than that for non-antibody loops. Provided that they belong to one of the canonical classes, very similar antibody loops are certain to exist in the database. Superposition of the stem residues for antibody loops also results in a better orientation than with arbitrary target loops because the neighboring residues tend to have a more similar beta-strand structure. Two H3 loops (for which no canonical structures have been proposed) were predicted with reasonable accuracy (RMSD of 0.49 A and 1.07 A) even though no corresponding antibody loops were in the database.

An interesting example of a structurally diverse group of sequentially homologous proteins is analyzed at the level of molecular interactions. In this family, the EF-hand calcium-binding proteins, there are examples of at least three distinct mutual positions of the N and C-terminal domains, despite significant sequence homology between all members of this family. Why does a particular protein choose one arrangement over another? To answer this question, detailed models of all proteins in their native structures as well as all alternative sequence/structure combinations are built by comparative modeling. By studying and comparing interactions stabilizing native structures and destabilizing alternative conformations, it is possible to gain insight into how such conformational diversity is achieved. It is shown that some mechanisms used to achieve it are: correlated mutations on the surface of two units and the presence of additional domains/chain fragments stabilizing desired topologies. The implications of these findings, both for structure predictions for other members of this family, as well as the general problem of quaternary structure formation, are discussed.

Grid-free protein folding simulations based on sequence and secondary structure knowledge (using mostly experimentally determined secondary structure information but also analysing results from secondary structure predictions) were investigated using the genetic algorithm, a backbone representation, and standard dihedral angular conformations. Optimal structures are selected according to basic protein building principles. Having previously applied this approach to proteins with helical topology, we have now developed additional criteria and weights for beta-strand-containing proteins, validated them on four small beta-strand-rich proteins with different topologies, and tested the general performance of the method on many further examples from known protein structures with mixed secondary structural type and less than 100 amino acid residues. Topology predictions close to the observed experimental structures were obtained in four test cases together with fitness values that correlated with the similarity of the predicted topology to the observed structures. Root-mean-square deviation values of C alpha atoms in the superposed predicted and observed structures, the latter of which had different topologies, were between 4.5 and 5.5 A(2.9 to 5.1 A without loops). Including 15 further protein examples with unique folds, root-mean-square deviation values ranged between 1.8 and 6.9 A with loop regions and averaged 5.3 A and 4.3 A, including and excluding loop regions, respectively.

We describe a comparative protein modelling method designed to find the most probable structure for a sequence given its alignment with related structures. The three-dimensional (3D) model is obtained by optimally satisfying spatial restraints derived from the alignment and expressed as probability density functions (pdfs) for the features restrained. For example, the probabilities for main-chain conformations of a modelled residue may be restrained by its residue type, main-chain conformation of an equivalent residue in a related protein, and the local similarity between the two sequences. Several such pdfs are obtained from the correlations between structural features in 17 families of homologous proteins which have been aligned on the basis of their 3D structures. The pdfs restrain C alpha-C alpha distances, main-chain N-O distances, main-chain and side-chain dihedral angles. A smoothing procedure is used in the derivation of these relationships to minimize the problem of a sparse database. The 3D model of a protein is obtained by optimization of the molecular pdf such that the model violates the input restraints as little as possible. The molecular pdf is derived as a combination of pdfs restraining individual spatial features of the whole molecule. The optimization procedure is a variable target function method that applies the conjugate gradients algorithm to positions of all non-hydrogen atoms. The method is automated and is illustrated by the modelling of trypsin from two other serine proteinases.

We introduce an approach to protein comparisons in which tertiary-structure information is exploited in the alignment of a protein sequence of known tertiary structure, or an aligned set of sequences of known homologous structures, with one or more sequences. The local tertiary environments of residues in the one or more three-dimensional structures (defined in terms of residue accessibility to solvent, secondary structure and hydrogen bonding) are used to select position-specific amino acid substitution scores and produce a scoring template suitable for aligning sequences or searching sequence data banks. The amino acid substitution scores have been accumulated from 72 families of protein structures in which the observed substitutions have been classified according to features of the local structure. Hence, the value attributed to a particular amino acid interchange in the template is not a constant, but is dependent upon the environmental context in which that substitution has occurred. We have used these structural templates to align proteins, as well as to search an amino acid sequence data bank for proteins having a similar fold. Indeed, a database of templates that corresponds to both unique structures and aligned homologous structures from the Brookhaven Protein Data Bank has been produced. A new sequence can be searched against the database of templates in order to identify a similar tertiary fold even if the sequence is not significantly similar to any proteins of known three-dimensional structure.

Protein homology modelling typically involves the prediction of side-chain conformations in the modelled protein while assuming a main-chain trace taken from a known tertiary structure of a protein with homologous sequence. It is generally believed that the need to examine all possible combinations of side-chain conformations poses the major obstacle to accurate homology modelling. Methods proposed heretofore use only discrete or limited searches of the side-chain torsion angle space to mitigate the combinatorial problem and also rely on simplified energy functions for calculational speed. The configurational constraints are typically based upon use of frequently observed torsion angles, fixed steps in torsion angles, or oligopeptide segments taken from tertiary structural databanks that are similar in sequence and conformation with the target structure. In the present work, a more fundamental approach is explored for several protein structures and it is demonstrated that the combinatorial barrier in side-chain placement hardly exists. Each side-group can be configured individually in the environment of only the backbone atoms using a systematic search procedure combined with extensive local energy minimization. Tests, using the main-chain or both the main-chain and remaining side-charm atoms to calculate low energy geometries for each residue, established the dominance of the main-chain contribution. The final structure is achieved by combining the individually placed side-chains followed by a full energy refinement of the structure. The prediction accuracy of the present homology modelling technique was assessed relative to other automated procedures and was found to yield improved predictions relative to the known side-chain conformations determined by X-ray crystallography.

Algorithms are suggested for identifying α-helical and β-structural regions in native globular proteins. α-Helical and β-structural regions are predicted, with accuracy of ~80 and ~85% respectively, for 25 proteins, the three-dimensional structures of which have been determined by X-ray diffraction crystallography. Secondary structure is predicted in 25 proteins with unknown three-dimensional structure.