Optimal prediction of underresolved dynamics

A method is presented for computing the average solution of problems that are too complicated for adequate resolution, but where information about the statistics of the solution is available. The method involves computing average derivatives by interpolation based on linear regression, and an updating of a measure constrained by the available crude information. Examples are given.

Prediction of operons in microbial genomes

Operon structure is an important organization feature of bacterial genomes. Many sets of genes occur in the same order on multiple genomes; these conserved gene groupings represent candidate operons. This study describes a computational method to estimate the likelihood that such conserved gene sets form operons. The method was used to analyze 34 bacterial and archaeal genomes, and yielded more than 7600 pairs of genes that are highly likely (P ≥ 0.98) to belong to the same operon. The sensitivity of our method is 30–50% for the Escherichia coli genome. The predicted gene pairs are available from our World Wide Web site http://www.tigr.org/tigr-scripts/operons/operons.cgi.

Statistical prediction of single-stranded regions in RNA secondary structure and application to predicting effective antisense target sites and beyond

Single-stranded regions in RNA secondary structure are important for RNA–RNA and RNA–protein interactions. We present a probability profile approach for the prediction of these regions based on a statistical algorithm for sampling RNA secondary structures. For the prediction of phylogenetically-determined single-stranded regions in secondary structures of representative RNA sequences, the probability profile offers substantial improvement over the minimum free energy structure. In designing antisense oligonucleotides, a practical problem is how to select a secondary structure for the target mRNA from the optimal structure(s) and many suboptimal structures with similar free energies. By summarizing the information from a statistical sample of probable secondary structures in a single plot, the probability profile not only presents a solution to this dilemma, but also reveals ‘well-determined’ single-stranded regions through the assignment of probabilities as measures of confidence in predictions. In antisense application to the rabbit β-globin mRNA, a significant correlation between hybridization potential predicted by the probability profile and the degree of inhibition of in vitro translation suggests that the probability profile approach is valuable for the identification of effective antisense target sites. Coupling computational design with DNA–RNA array technique provides a rational, efficient framework for antisense oligonucleotide screening. This framework has the potential for high-throughput applications to functional genomics and drug target validation.

Recent improvements in prediction of protein structure by global optimization of a potential energy function

Recent improvements of a hierarchical ab initio or de novo approach for predicting both α and β structures of proteins are described. The united-residue energy function used in this procedure includes multibody interactions from a cumulant expansion of the free energy of polypeptide chains, with their relative weights determined by Z-score optimization. The critical initial stage of the hierarchical procedure involves a search of conformational space by the conformational space annealing (CSA) method, followed by optimization of an all-atom model. The procedure was assessed in a recent blind test of protein structure prediction (CASP4). The resulting lowest-energy structures of the target proteins (ranging in size from 70 to 244 residues) agreed with the experimental structures in many respects. The entire experimental structure of a cyclic α-helical protein of 70 residues was predicted to within 4.3 Å α-carbon (Cα) rms deviation (rmsd) whereas, for other α-helical proteins, fragments of roughly 60 residues were predicted to within 6.0 Å Cα rmsd. Whereas β structures can now be predicted with the new procedure, the success rate for α/β- and β-proteins is lower than that for α-proteins at present. For the β portions of α/β structures, the Cα rmsd's are less than 6.0 Å for contiguous fragments of 30–40 residues; for one target, three fragments (of length 10, 23, and 28 residues, respectively) formed a compact part of the tertiary structure with a Cα rmsd less than 6.0 Å. Overall, these results constitute an important step toward the ab initio prediction of protein structure solely from the amino acid sequence.

Prediction of protein deamidation rates from primary and three-dimensional structure

A method for the quantitative estimation of instability with respect to deamidation of the asparaginyl (Asn) residues in proteins is described. The procedure involves the observation of several simple aspects of the three-dimensional environment of each Asn residue in the protein and a calculation that includes these observations, the primary amino acid residue sequence, and the previously reported complete set of sequence-dependent rates of deamidation for Asn pentapeptides. This method is demonstrated and evaluated for 23 proteins in which 31 unstable and 167 stable Asn residues have been reported and for 7 unstable and 63 stable Asn residues that have been reported in 61 human hemoglobin variants. The relative importance of primary structure and three-dimensional structure in Asn deamidation is estimated.

Prediction of the stability of DNA triplexes.

We present rules that allow one to predict the stability of DNA pyrimidine.purine.pyrimidine (Y.R.Y) triple helices on the basis of the sequence. The rules were derived from van't Hoff analysis of 23 oligonucleotide triplexes tested at a variety of pH values. To predict the enthalpy of triplex formation (delta H degrees), a simple nearest-neighbor model was found to be sufficient. However, to accurately predict the free energy of the triplex (delta G degrees), a combination model consisting of five parameters was needed. These parameters were (i) the delta G degrees for helix initiation, (ii) the delta G degrees for adding a T-A.T triple, (iii) the delta G degrees for adding a C(+)-G.C triple, (iv) the penalty for adjacent C bases, and (v) the pH dependence of the C(+)-G.C triple's stability. The fitted parameters are highly consistent with thermodynamic data from the basis set, generally predicting both delta H degrees and delta G degrees to within the experimental error. Examination of the parameters points out several interesting features. The combination model predicts that C(+) -G.C. triples are much more stabilizing than T-A.T triples below pH 7.0 and that the stability of the former increases approximately equal to 1 kcal/mol per pH unit as the pH is decreased. Surprisingly though, the most stable sequence is predicted to be a CT repeat, as adjacent C bases partially cancel the stability of one another. The parameters successfully predict tm values from other laboratories, with some interesting exceptions.

Definition of the HLA-A29 peptide ligand motif allows prediction of potential T-cell epitopes from the retinal soluble antigen, a candidate autoantigen in birdshot retinopathy.

The peptide-binding motif of HLA-A29, the predisposing allele for birdshot retinopathy, was determined after acid-elution of endogenous peptides from purified HLA-A29 molecules. Individual and pooled HPLC fractions were sequenced by Edman degradation. Major anchor residues could be defined as glutamate at the second position of the peptide and as tyrosine at the carboxyl terminus. In vitro binding of polyglycine synthetic peptides to purified HLA-A29 molecules also revealed the need for an auxiliary anchor residue at the third position, preferably phenylalanine. By using this motif, we synthesized six peptides from the retinal soluble antigen, a candidate autoantigen in autoimmune uveoretinitis. Their in vitro binding was tested on HLA-A29 and also on HLA-B44 and HLA-B61, two alleles sharing close peptide-binding motifs. Two peptides derived from the carboxyl-terminal sequence of the human retinal soluble antigen bound efficiently to HLA-A29. This study could contribute to the prediction of T-cell epitopes from retinal autoantigens implicated in birdshot retinopathy.

The invariant system of coordinates of antibody molecules: prediction of the "standard" C alpha framework of VL and VH domains.

A new approach of comparing protein structures that does not involve the procedure of superposition is suggested. An invariant system of coordinates for immunoglobulin molecules that is based on the geometrical symmetry inherent to the variable domain light-chain (VL)-heavy-chain (VH) complex is described. The coordinates of the Calpha atoms in 22 immunoglobulin structures are calculated in the invariant system of coordinates. We found that 76 identical positions in this Calpha framework are symmetrical about the twofold axis. Comparison of the identical positions in these molecules allows us to select 96 positions in the light chains and 87 positions in the heavy chains whose Calpha atom coordinates are approximately the same. To check whether the average coordinates of Calpha atoms in these positions complies with the stereochemical requirements, we calculated Calpha-Calpha distances. Seventy-three positions of the light chains and 72 positions of the heavy chains satisfy the Calpha-Calpha distance criterion. The Calpha atoms in these positions are used for constructing the "standard" Calpha framework of VL and VH complexes. The average coordinates of Calpha atoms are presented.

Theoretical prediction of a crystal structure.

The diffusion equation method of global minimization is applied to compute the crystal structure of S6, with no a priori knowledge about the system. The experimental lattice parameters and positions and orientations of the molecules in the unit cell are predicted correctly.

Prediction of protein folding class using global description of amino acid sequence.

We present a method for predicting protein folding class based on global protein chain description and a voting process. Selection of the best descriptors was achieved by a computer-simulated neural network trained on a data base consisting of 83 folding classes. Protein-chain descriptors include overall composition, transition, and distribution of amino acid attributes, such as relative hydrophobicity, predicted secondary structure, and predicted solvent exposure. Cross-validation testing was performed on 15 of the largest classes. The test shows that proteins were assigned to the correct class (correct positive prediction) with an average accuracy of 71.7%, whereas the inverse prediction of proteins as not belonging to a particular class (correct negative prediction) was 90-95% accurate. When tested on 254 structures used in this study, the top two predictions contained the correct class in 91% of the cases.