Prediction of protein functional residues from sequence by probability density estimation

MOTIVATION: The prediction of ligand-binding residues or catalytically active residues of a protein may give important hints that can guide further genetic or biochemical studies. Existing sequence-based prediction methods mostly rank residue positions by evolutionary conservation calculated from a multiple sequence alignment of homologs. A problem hampering more wide-spread application of these methods is the low per-residue precision, which at 20% sensitivity is around 35% for ligand-binding residues and 20% for catalytic residues. RESULTS: We combine information from the conservation at each site, its amino acid distribution, as well as its predicted secondary structure (ss) and relative solvent accessibility (rsa). First, we measure conservation by how much the amino acid distribution at each site differs from the distribution expected for the predicted ss and rsa states. Second, we include the conservation of neighboring residues in a weighted linear score by analytically optimizing the signal-to-noise ratio of the total score. Third, we use conditional probability density estimation to calculate the probability of each site to be functional given its conservation, the observed amino acid distribution, and the predicted ss and rsa states. We have constructed two large data sets, one based on the Catalytic Site Atlas and the other on PDB SITE records, to benchmark methods for predicting functional residues. The new method FRcons predicts ligand-binding and catalytic residues with higher precision than alternative methods over the entire sensitivity range, reaching 50% and 40% precision at 20% sensitivity, respectively. AVAILABILITY: Server: http://frpred.tuebingen.mpg.de. Data sets: ftp://ftp.tuebingen.mpg.de/pub/protevo/FRpred/.     

Prediction of protein function from protein sequence and structure

The sequence of a genome contains the plans of the possible life of an organism, but implementation of genetic information depends on the functions of the proteins and nucleic acids that it encodes. Many individual proteins of known sequence and structure present challenges to the understanding of their function. In particular, a number of genes responsible for diseases have been identified but their specific functions are unknown. Whole-genome sequencing projects are a major source of proteins of unknown function. Annotation of a genome involves assignment of functions to gene products, in most cases on the basis of amino-acid sequence alone. 3D structure can aid the assignment of function, motivating the challenge of structural genomics projects to make structural information available for novel uncharacterized proteins. Structure-based identification of homologues often succeeds where sequence-alone-based methods fail, because in many cases evolution retains the folding pattern long after sequence similarity becomes undetectable. Nevertheless, prediction of protein function from sequence and structure is a difficult problem, because homologous proteins often have different functions. Many methods of function prediction rely on identifying similarity in sequence and/or structure between a protein of unknown function and one or more well-understood proteins. Alternative methods include inferring conservation patterns in members of a functionally uncharacterized family for which many sequences and structures are known. However, these inferences are tenuous. Such methods provide reasonable guesses at function, but are far from foolproof. It is therefore fortunate that the development of whole-organism approaches and comparative genomics permits other approaches to function prediction when the data are available. These include the use of protein-protein interaction patterns, and correlations between occurrences of related proteins in different organisms, as indicators of functional properties. Even if it is possible to ascribe a particular function to a gene product, the protein may have multiple functions. A fundamental problem is that function is in many cases an ill-defined concept. In this article we review the state of the art in function prediction and describe some of the underlying difficulties and successes.     

Prediction of protein domain boundaries from sequence alone

We present here a simple approach to identify domain boundaries in proteins of an unknown three-dimensional structure. Our method is based on the hypothesis that a high-side chain entropy of a region in a protein chain must be compensated by a high-residue interaction energy within the region, which could correlate with a well-structured part of the globule, that is, with a domain unit. For protein domains, this means that the domain boundary is conditioned by amino acid residues with a small value of side chain entropy, which correlates with the side chain size. On the one hand, relatively high Ala and Gly content on the domain boundary results in high conformational entropy of the backbone chain between the domains. On the other hand, the presence of Pro residues leads to the formation of hinges for a relative orientation of domains. The method was applied to 646 proteins with two contiguous domains extracted from the SCOP database with a success rate of 63%. We also report the prediction of domain boundaries for CASP5 targets obtained with the same method.     

Prediction of protein secondary structure from amino acid sequence

The conformational parametersP _k for each amino acid species (j=1–20) of sequential peptides in proteins are presented as the product ofP _i,k , wherei is the number of the sequential residues in thekth conformational state (k=-helix,-sheet,-turn, or unordered structure). Since the average parameter for ann-residue segment is related to the average probability of finding the segment in the kth state, it becomes a geometric mean of (P _k )_av=(P _i,k ) ^1/n with amino acid residuei increasing from 1 ton. We then used ln(P_k)_av to convert a multiplicative process to a summation, i.e., ln(P _k ) _av =(1/n)P _i,k (i=1 ton) for ease of operation. However, this is unlike the popular Chou-Fasman algorithm, which has the flaw of using the arithmetic mean for relative probabilities. The Chou-Fasman algorithm happens to be close to our calculations in many cases mainly because the difference between theirP _k and our InP _k is nearly constant for about one-half of the 20 amino acids. When stronger conformation formers and breakers exist, the difference become larger and the prediction at the N- and C-terminal-helix or-sheet could differ. If the average conformational parameters of the overlapping segments of any two states are too close for a unique solution, our calculations could lead to a different prediction.     

Prediction of Saccharomyces cerevisiae protein functional class from functional domain composition

MOTIVATION: A key goal of genomics is to assign function to genes, especially for orphan sequences. RESULTS: We compared the clustered functional domains in the SBASE database to each protein sequence using BLASTP. This representation for a protein is a vector, where each of the non-zero entries in the vector indicates a significant match between the sequence of interest and the SBASE domain. The machine learning methods nearest neighbour algorithm (NNA) and support vector machines are used for predicting protein functional classes from this information. We find that the best results are found using the SBASE-A database and the NNA, namely 72% accuracy for 79% coverage. We tested an assigning function based on searching for InterPro sequence motifs and by taking the most significant BLAST match within the dataset. We applied the functional domain composition method to predict the functional class of 2018 currently unclassified yeast open reading frames. AVAILABILITY: A program for the prediction method, that uses NNA called Functional Class Prediction based on Functional Domains (FCPFD) is available and can be obtained by contacting Y.D.Cai at y.cai@umist.ac.uk     

Molecular and functional characterization of a unique Rab protein,RABRP1, containing the WDIAGQE sequence in a GTPase motif

Rab proteins of the small G-protein superfamily are known to be involved in intracellular vesicle transport. Here, we describe the unique characteristics of a novel Rab protein, RABRP1 (Rab-Related Protein 1). The Drosophila RabRP1 gene is mainly transcribed in the eyes and testes, where the 3-kb and 1.5-kb mRNAs, respectively, are the predominant gene products. The amino-acid sequence deduced from the longer cDNA indicated that the C-terminal 1/3 of the sequence shares homology with Rab proteins, whereas the rest of the peptide shows no significant homology with any other proteins. Immunoblot analysis using antiserum against the Rab-domain indicated that the multiple translates (94 k, 53 k, 30 k, 29 k and 27 k) were expressed in the eyes. In contrast, only smaller peptides (30 k, 29 k and 27 k) were identified in the testes. Molecular phylogenetic analysis revealed that RABRP1 forms a subgroup with Dictiostelium RabE and mammalian Rab29, Rab32, Rab38 proteins, whose functions have not been identified yet. RABRP1 and its relatives were characterized by the amino acid substitution occurring in the conserved GTP-binding motifs. Immunohistochemical studies demonstrated that RABRP1 was localized on the subrhabdomeric cisternae of photoreceptor cells and on the pigment granules in photoreceptor and pigment cells in the retina. The expression of the dominant negative RABRP1 caused the abnormal accumulation of autophagosome-like vesicles. These data suggest that RABRP1 is involved in the lysosomal vesicle transport pathway, including the biogenesis or degradation of pigment granules.     

The elucidation of protein function by sequence motif analysis

Protein sequence motifs are acquiring increasing prominencein the area of sequence analysis. This review describes thecurrent methods of their construction and their use in the determinationof protein function, and offers guidelines on interpreting dataobtained. An appendix is attached which refers to 200 motifsof various kinds.     

Prediction of protein structural class from the amino acid sequence

The multidimensional statistical technique of discriminant analysis is used to allocate amino acid sequences to one of four secondary structural classes: high α content, high β content, mixed α and β, low content of ordered structure. Discrimination is based on four attributes: estimates of percentages of α and β structures, and regular variations in the hydrophobic values of residues along the sequence, occurring with periods of 2 and 3.6 residues. The reliability of the method, estimated by classifying 138 sequences from the Brookhaven Protein Data Bank, is 80%, with no misallocations between α-rich and β-rich classes. The reliability can be increased to 84% by making no allocation for proteins classified with odds close to 1. Classification using previously developed secondary structural prediction methods is considerably less reliable, the best result being 64% obtained using predictions based on the Delphi method.     

Prediction of functional tertiary interactions and intermolecular interfaces from primary sequence data

Given the availability of sequence information for many species, one can examine how the sequence of a gene varies among different organisms. This is accomplished by aligning the sequences and observing patterns of conservation, mutation and counter-mutation at different positions in the gene. Imbedded in these patterns is information on energetic coupling and macromolecular interactions, which can be deciphered by application of statistical algorithms. Here we report a robust approach for predicting interactions within (or between) any type of biopolymer, including proteins, RNAs and RNA-protein complexes. Rather than maximize the number of predictions, this approach is designed to detect a limited number of highly significant interactions, thereby providing accurate results from alignments that contain a modest number of sequences (20-60). The versatility and accuracy of the algorithm is demonstrated by the successful prediction of important intramolecular interactions within RNAs, modified RNAs, and proteins, as well as the prediction of RNA-protein and protein-protein interactions.     

Analysis and prediction of functional sub-types from protein sequence alignments

The increasing number and diversity of protein sequence families requires new methods to define and predict details regarding function. Here, we present a method for analysis and prediction of functional sub-types from multiple protein sequence alignments. Given an alignment and set of proteins grouped into sub-types according to some definition of function, such as enzymatic specificity, the method identifies positions that are indicative of functional differences by comparison of sub-type specific sequence profiles, and analysis of positional entropy in the alignment. Alignment positions with significantly high positional relative entropy correlate with those known to be involved in defining sub-types for nucleotidyl cyclases, protein kinases, lactate/malate dehydrogenases and trypsin-like serine proteases. We highlight new positions for these proteins that suggest additional experiments to elucidate the basis of specificity. The method is also able to predict sub-type for unclassified sequences. We assess several variations on a prediction method, and compare them to simple sequence comparisons. For assessment, we remove close homologues to the sequence for which a prediction is to be made (by a sequence identity above a threshold). This simulates situations where a protein is known to belong to a protein family, but is not a close relative of another protein of known sub-type. Considering the four families above, and a sequence identity threshold of 30 %, our best method gives an accuracy of 96 % compared to 80 % obtained for sequence similarity and 74 % for BLAST. We describe the derivation of a set of sub-type groupings derived from an automated parsing of alignments from PFAM and the SWISSPROT database, and use this to perform a large-scale assessment. The best method gives an average accuracy of 94 % compared to 68 % for sequence similarity and 79 % for BLAST. We discuss implications for experimental design, genome annotation and the prediction of protein function and protein intra-residue distances.     

Prediction of loop regions in protein sequence

We suggest an algorithm that inputs a protein sequence and outputs a decomposition of the protein chain into a regular part including secondary structures and a nonregular part corresponding to loop regions. We have analyzed loop regions in a protein dataset of 3,769 globular domains and defined the optimal parameters for this prediction: the threshold between regular and nonregular regions and the optimal window size for averaging procedures using the scale of the expected number of contacts in a globular state and entropy scale as the number of degrees of freedom for the angles phi, psi, and chi for each amino acid. Comparison with known methods demonstrates that our method gives the same results as the well-known ALB method based on physical properties of amino acids (the percentage of true predictions is 64% against 66%), and worse prediction for regular and nonregular regions than PSIPRED (Protein Structure Prediction Server) without alignment of homologous proteins (the percentage of true predictions is 73%). The potential advantage of the suggested approach is that the predicted set of loops can be used to find patterns of rigid and flexible loops as possible candidates to play a structure/function role as well as a role of antigenic determinants.     

Thirty-plus functional families from a single motif

It is now possible to identify over 30 functional subfamilies among the WD-repeat-containing proteins found in the completed genomes. The majority of these subfamilies have at least one member for which experimental data allow assignment to a cellular pathway or process. Half of the 63 WD-repeat-containing proteins in Saccharomyces cerevisiae, half of the 70 in Caenorhabditis elegans, and a third of the 100 plus predicted in Drosophila can be assigned to 23 of these functional subfamilies. Perhaps indicative of the future, 33 WD-repeat-containing proteins from the partial genome of Arabidopsis thaliana can now be assigned to 18 of these subfamilies. These assignments have been made possible by combining traditional sequence similarity with an implied common beta propeller structural context to obtain measures of protein-protein surface similarity. The beta propeller structural context is represented in the form of a Hidden Markov Model. The procedure is completely automated.     

Canonical protein splicing of a class 1 intein that has a class 3 noncanonical sequence motif

A Thermobifida fusca intein has two characteristics of class 3 inteins: a noncontiguous covariant Trp-Cys-Thr triplet and a Ser flanking its C terminus. However, it has Cys at position one, characteristic of class 1 inteins. Splicing does not require the internal Cys, which may instead coordinate the active site. Therefore, the intein is class 1.     

Prediction of solvent accessibility and sites of deleterious mutations from protein sequence

Residues that form the hydrophobic core of a protein are critical for its stability. A number of approaches have been developed to classify residues as buried or exposed. In order to optimize the classification, we have refined a suite of five methods over a large dataset and proposed a metamethod based on an ensemble average of the individual methods, leading to a two-state classification accuracy of 80%. Many studies have suggested that hydrophobic core residues are likely sites of deleterious mutations, so we wanted to see to what extent these sites can be predicted from the putative buried residues. Residues that were most confidently classified as buried were proposed as sites of deleterious mutations. This proposition was tested on six proteins for which sites of deleterious mutations have previously been identified by stability measurement or functional assay. Of the total of 130 residues predicted as sites of deleterious mutations, 104 (or 80%) were correct.     

Prediction of protein helix content from an autocorrelation analysis of sequence hydrophobicities

It is demonstrated that protein α-helix content can be predicted from an autocorrelation analysis of the protein hydrophobicity sequence. The Fourier transform of the autocorrelation function yields the spectral densities or weights of the various frequencies contributing to the autocorrelation function. Using sequence and secondary structure data from more than 160 proteins and domains, a linear relationship was found between spectral density at periodicity 3.7 and protein α-helix content (r = 0.83). This relation permits prediction of the helix content (x) of proteins of known sequence to within ± 15%, i.e., as (x ± 15)%. Predictions based on the autocorrelation procedure are compared with values obtained by other methods.     

Prediction of functional specificity determinants from protein sequences using log-likelihood ratios

MOTIVATION: A number of methods have been developed to predict functional specificity determinants in protein families based on sequence information. Most of these methods rely on pre-defined functional subgroups. Manual subgroup definition is difficult because of the limited number of experimentally characterized subfamilies with differing specificity, while automatic subgroup partitioning using computational tools is a non-trivial task and does not always yield ideal results. RESULTS: We propose a new approach SPEL (specificity positions by evolutionary likelihood) to detect positions that are likely to be functional specificity determinants. SPEL, which does not require subgroup definition, takes a multiple sequence alignment of a protein family as the only input, and assigns a P-value to every position in the alignment. Positions with low P-values are likely to be important for functional specificity. An evolutionary tree is reconstructed during the calculation, and P-value estimation is based on a random model that involves evolutionary simulations. Evolutionary log-likelihood is chosen as a measure of amino acid distribution at a position. To illustrate the performance of the method, we carried out a detailed analysis of two protein families (LacI/PurR and G protein alpha subunit), and compared our method with two existing methods (evolutionary trace and mutual information based). All three methods were also compared on a set of protein families with known ligand-bound structures. AVAILABILITY: SPEL is freely available for non-commercial use. Its pre-compiled versions for several platforms and alignments used in this work are available at ftp://iole.swmed.edu/pub/SPEL/     

Prediction of heme binding residues from protein sequences with integrative sequence profiles

Background

The heme-protein interactions are essential for various biological processes such as electron transfer, catalysis, signal transduction and the control of gene expression. The knowledge of heme binding residues can provide crucial clues to understand these activities and aid in functional annotation, however, insufficient work has been done on the research of heme binding residues from protein sequence information.

Methods

We propose a sequence-based approach for accurate prediction of heme binding residues by a novel integrative sequence profile coupling position specific scoring matrices with heme specific physicochemical properties. In order to select the informative physicochemical properties, we design an intuitive feature selection scheme by combining a greedy strategy with correlation analysis.

Results

Our integrative sequence profile approach for prediction of heme binding residues outperforms the conventional methods using amino acid and evolutionary information on the 5-fold cross validation and the independent tests.

Conclusions

The novel feature of an integrative sequence profile achieves good performance using a reduced set of feature vector elements.

     

Prediction of protein folding from amino acid sequence over discrete conformation spaces

Predicting the three-dimensional structure of a protein given only its amino acid sequence is a long-standing goal in computational chemistry. In the thermodynamic approach, one needs a potential function of conformation that resembles the free energy of the real protein to the extent that the global minimum of the potential is attained by the native conformation and no other. In practice, this has never been achieved with certainty because even with greatly simplified representations of the polypeptide chain, there are an astronomical number of local minima to examine. If one chooses instead a protein representation with only a large but manageable number of discrete conformations, then the global preference of the potential for the native can be directly verified. Representing a protein as a walk on a two-dimensional square lattice makes it easy to see that simple functions of the interresidue contacts are sufficient to globally favor a given "native" conformation, as long as it is a compact, globular structure. Explicit representation of the solvent is not required. Another more realistic way to confine the conformational search to a finite set is to draw alternative conformations from fragments of larger proteins having known crystal structure. Then it is possible to construct a simple function of interresidue contacts in three dimensions such that only 8 proteins are required to determine the adjustable parameters, and the native conformations of 37 other proteins are correctly preferred over all alternative conformations. The deduced function favors short-range backbone-backbone contacts regardless of residue type and long-range hydrophobic associations. Interactions over long distances, such as electrostatics, are not required.