Combining algorithms to predict bacterial protein sub-cellular location: Parallel versus concurrent implementations

We describe a novel and potentially important tool for candidate subunit vaccine selection through in silico reverse-vaccinology. A set of Bayesian networks able to make individual predictions for specific subcellular locations is implemented in three pipelines with different architectures: a parallel implementation with a confidence level-based decision engine and two serial implementations with a hierarchical decision structure, one initially rooted by prediction between membrane types and another rooted by soluble versus membrane prediction. The parallel pipeline outperformed the serial pipeline, but took twice as long to execute. The soluble-rooted serial pipeline outperformed the membrane-rooted predictor. Assessment using genomic test sets was more equivocal, as many more predictions are made by the parallel pipeline, yet the serial pipeline identifies 22 more of the 74 proteins of known location.     

Multi-class subcellular location prediction for bacterial proteins

Two algorithms, based onBayesian Networks (BNs), for bacterial subcellular location prediction, are explored in this paper: one predicts all locations for Gram+ bacteria and the other all locations for Gram- bacteria. Methods were evaluated using different numbers of residues (from the N-terminal 10 residues to the whole sequence) and residue representation (amino acid-composition, percentage amino acid-composition or normalised amino acid-composition). The accuracy of the best resulting BN was compared to PSORTB. The accuracy of this multi-location BN was roughly comparable to PSORTB; the difference in predictions is low, often less than 2%. The BN method thus represents both an important new avenue of methodological development for subcellular location prediction and a potentially value new tool of true utilitarian value for candidate subunit vaccine selection.     

Functional annotation prediction: all for one and one for all

In an era of rapid genome sequencing and high-throughput technology, automatic function prediction for a novel sequence is of utter importance in bioinformatics. While automatic annotation methods based on local alignment searches can be simple and straightforward, they suffer from several drawbacks, including relatively low sensitivity and assignment of incorrect annotations that are not associated with the region of similarity. ProtoNet is a hierarchical organization of the protein sequences in the UniProt database. Although the hierarchy is constructed in an unsupervised automatic manner, it has been shown to be coherent with several biological data sources. We extend the ProtoNet system in order to assign functional annotations automatically. By leveraging on the scaffold of the hierarchical classification, the method is able to overcome some frequent annotation pitfalls.     

Euk-PLoc: an ensemble classifier for large-scale eukaryotic protein subcellular location prediction

With the avalanche of newly-found protein sequences emerging in the post genomic era, it is highly desirable to develop an automated method for fast and reliably identifying their subcellular locations because knowledge thus obtained can provide key clues for revealing their functions and understanding how they interact with each other in cellular networking. However, predicting subcellular location of eukaryotic proteins is a challenging problem, particularly when unknown query proteins do not have significant homology to proteins of known subcellular locations and when more locations need to be covered. To cope with the challenge, protein samples are formulated by hybridizing the information derived from the gene ontology database and amphiphilic pseudo amino acid composition. Based on such a representation, a novel ensemble hybridization classifier was developed by fusing many basic individual classifiers through a voting system. Each of these basic classifiers was engineered by the KNN (K-Nearest Neighbor) principle. As a demonstration, a new benchmark dataset was constructed that covers the following 18 localizations: (1) cell wall, (2) centriole, (3) chloroplast, (4) cyanelle, (5) cytoplasm, (6) cytoskeleton, (7) endoplasmic reticulum, (8) extracell, (9) Golgi apparatus, (10) hydrogenosome, (11) lysosome, (12) mitochondria, (13) nucleus, (14) peroxisome, (15) plasma membrane, (16) plastid, (17) spindle pole body, and (18) vacuole. To avoid the homology bias, none of the proteins included has > or =25% sequence identity to any other in a same subcellular location. The overall success rates thus obtained via the 5-fold and jackknife cross-validation tests were 81.6 and 80.3%, respectively, which were 40-50% higher than those performed by the other existing methods on the same strict dataset. The powerful predictor, named "Euk-PLoc", is available as a web-server at http://202.120.37.186/bioinf/euk . Furthermore, to support the need of people working in the relevant areas, a downloadable file will be provided at the same website to list the results predicted by Euk-PLoc for all eukaryotic protein entries (excluding fragments) in Swiss-Prot database that do not have subcellular location annotations or are annotated as being uncertain. The large-scale results will be updated twice a year to include the new entries of eukaryotic proteins and reflect the continuous development of Euk-PLoc.     

LOC3D: annotate sub-cellular localization for protein structures

LOC3D (http://cubic.bioc.columbia.edu/db/LOC3d/) is both a weekly-updated database and a web server for predictions of sub-cellular localization for eukaryotic proteins of known three-dimensional (3D) structure. Localization is predicted using four different methods: (i) PredictNLS, prediction of nuclear proteins through nuclear localization signals; (ii) LOChom, inferring localization through sequence homology; (iii) LOCkey, inferring localization through automatic text analysis of SWISS-PROT keywords; and (iv) LOC3Dini, ab initio prediction through a system of neural networks and vector support machines. The final prediction is based on the method that predicts localization with the highest confidence. The LOC3D database currently contains predictions for >8700 eukaryotic protein chains taken from the Protein Data Bank (PDB). The web server can be used to predict sub-cellular localization for proteins for which only a predicted structure is available from threading servers. This makes the resource of particular interest to structural genomics initiatives.     

SOSUI-GramN: high performance prediction for sub-cellular localization of proteins in gram-negative bacteria

A predictive software system, SOSUI-GramN, was developed for assessing the subcellular localization of proteins in Gram-negative bacteria. The system does not require the sequence homology data of any known sequences; instead, it uses only physicochemical parameters of the N- and C-terminal signal sequences, and the total sequence. The precision of the prediction system for subcellular localization to extracellular, outer membrane, periplasm, inner membrane and cytoplasmic medium was 92.3%, 89.4%, 86.4%, 97.5% and 93.5%, respectively, with corresponding recall rates of 70.3%, 87.5%, 76.0%, 97.5% and 88.4%, respectively. The overall performance for precision and recall obtained using this method was 92.9% and 86.7%, respectively. The comparison of performance of SOSUI-GramN with that of other methods showed the performance of prediction for extracellular proteins, as well as inner and outer membrane proteins, was either superior or equivalent to that obtained with other systems. SOSUI-GramN particularly improved the accuracy for predictions of extracellular proteins which is an area of weakness common to the other methods.     

Toward minimalist models of larger proteins: a ubiquitin-like protein

Our recently developed off-lattice bead model capable of simulating protein structures with mixed alpha/beta content has been extended to model the folding of a ubiquitin-like protein and provides a means for examining the more complex kinetics involved in the folding of larger proteins. Using trajectories generated from constant-temperature Langevin dynamics simulations and sampling with the multiple multi-histogram method over five-order parameters, we are able to characterize the free energy landscape for folding and find evidence for folding through compact intermediates. Our model reproduces the observation that the C-terminus loop structure in ubiquitin is the last to fold in the folding process and most likely plays a spectator role in the folding kinetics. The possibility of a productive metastable intermediate along the folding pathway consisting of collapsed states with no secondary structure, and of intermediates or transition structures involving secondary structural elements occurring early in the sequence, is also supported by our model. The kinetics of folding remain multi-exponential below the folding temperature, with glass-like kinetics appearing at T/T(f) approximately 0.86. This new physicochemical model, designed to be predictive, helps validate the value of modeling protein folding at this level of detail for genomic-scale studies, and motivates further studies of other protein topologies and the impact of more complex energy functions, such as the addition of solvation forces.     

Construction of two mathematical models for prediction of bacterial sRNA targets

The signal recognition particle (SRP) mediated protein translocation pathway is universal and highly conserved in all kingdoms of life. Significant progresses have been made to understand its molecular mechanism, yet many open questions remain. A structure model, showing how nascent peptide inserts into peptide translocon with the help of SRP protein Ffh and its receptor FtsY, is desired to facilitate our studies. In this work, we presented such a model derived by computational docking of the Ffh-FtsY complex onto the translocon. This model was compatible with most available experiments. It suggested that the Ffh-FtsY complex approached the translocon with its G domains and was locked up by the cytoplasmic loop of SecG and the C5/C6 loops of SecY. Several residues were expected to play important roles in regulating GTP hydrolysis. Additionally, a hypothesis on the yet ambiguous function of FtsY A domain was proposed. These interesting results invite experimental investigations.     

Diffusion kernel-based logistic regression models for protein function prediction

Assigning functions to unknown proteins is one of the most important problems in proteomics. Several approaches have used protein-protein interaction data to predict protein functions. We previously developed a Markov random field (MRF) based method to infer a protein's functions using protein-protein interaction data and the functional annotations of its protein interaction partners. In the original model, only direct interactions were considered and each function was considered separately. In this study, we develop a new model which extends direct interactions to all neighboring proteins, and one function to multiple functions. The goal is to understand a protein's function based on information on all the neighboring proteins in the interaction network. We first developed a novel kernel logistic regression (KLR) method based on diffusion kernels for protein interaction networks. The diffusion kernels provide means to incorporate all neighbors of proteins in the network. Second, we identified a set of functions that are highly correlated with the function of interest, referred to as the correlated functions, using the chi-square test. Third, the correlated functions were incorporated into our new KLR model. Fourth, we extended our model by incorporating multiple biological data sources such as protein domains, protein complexes, and gene expressions by converting them into networks. We showed that the KLR approach of incorporating all protein neighbors significantly improved the accuracy of protein function predictions over the MRF model. The incorporation of multiple data sets also improved prediction accuracy. The prediction accuracy is comparable to another protein function classifier based on the support vector machine (SVM), using a diffusion kernel. The advantages of the KLR model include its simplicity as well as its ability to explore the contribution of neighbors to the functions of proteins of interest.     

Using functional domain composition and support vector machines for prediction of protein subcellular location

Proteins are generally classified into the following 12 subcellular locations: 1) chloroplast, 2) cytoplasm, 3) cytoskeleton, 4) endoplasmic reticulum, 5) extracellular, 6) Golgi apparatus, 7) lysosome, 8) mitochondria, 9) nucleus, 10) peroxisome, 11) plasma membrane, and 12) vacuole. Because the function of a protein is closely correlated with its subcellular location, with the rapid increase in new protein sequences entering into databanks, it is vitally important for both basic research and pharmaceutical industry to establish a high throughput tool for predicting protein subcellular location. In this paper, a new concept, the so-called "functional domain composition" is introduced. Based on the novel concept, the representation for a protein can be defined as a vector in a high-dimensional space, where each of the clustered functional domains derived from the protein universe serves as a vector base. With such a novel representation for a protein, the support vector machine (SVM) algorithm is introduced for predicting protein subcellular location. High success rates are obtained by the self-consistency test, jackknife test, and independent dataset test, respectively. The current approach not only can play an important complementary role to the powerful covariant discriminant algorithm based on the pseudo amino acid composition representation (Chou, K. C. (2001) Proteins Struct. Funct. Genet. 43, 246-255; Correction (2001) Proteins Struct. Funct. Genet. 44, 60), but also may greatly stimulate the development of this area.     

Euk-mPLoc: a fusion classifier for large-scale eukaryotic protein subcellular location prediction by incorporating multiple sites

One of the critical challenges in predicting protein subcellular localization is how to deal with the case of multiple location sites. Unfortunately, so far, no efforts have been made in this regard except for the one focused on the proteins in budding yeast only. For most existing predictors, the multiple-site proteins are either excluded from consideration or assumed even not existing. Actually, proteins may simultaneously exist at, or move between, two or more different subcellular locations. For instance, according to the Swiss-Prot database (version 50.7, released 19-Sept-2006), among the 33,925 eukaryotic protein entries that have experimentally observed subcellular location annotations, 2715 have multiple location sites, meaning about 8% bearing the multiplex feature. Proteins with multiple locations or dynamic feature of this kind are particularly interesting because they may have some very special biological functions intriguing to investigators in both basic research and drug discovery. Meanwhile, according to the same Swiss-Prot database, the number of total eukaryotic protein entries (except those annotated with "fragment" or those with less than 50 amino acids) is 90,909, meaning a gap of (90,909-33,925) = 56,984 entries for which no knowledge is available about their subcellular locations. Although one can use the computational approach to predict the desired information for the blank, so far, all the existing methods for predicting eukaryotic protein subcellular localization are limited in the case of single location site only. To overcome such a barrier, a new ensemble classifier, named Euk-mPLoc, was developed that can be used to deal with the case of multiple location sites as well. Euk-mPLoc is freely accessible to the public as a Web server at http://202.120.37.186/bioinf/euk-multi. Meanwhile, to support the people working in the relevant areas, Euk-mPLoc has been used to identify all eukaryotic protein entries in the Swiss-Prot database that do not have subcellular location annotations or are annotated as being uncertain. The large-scale results thus obtained have been deposited at the same Web site via a downloadable file prepared with Microsoft Excel and named "Tab_Euk-mPLoc.xls". Furthermore, to include new entries of eukaryotic proteins and reflect the continuous development of Euk-mPLoc in both the coverage scope and prediction accuracy, we will timely update the downloadable file as well as the predictor, and keep users informed by publishing a short note in the Journal and making an announcement in the Web Page.     

Root location in random trees: A polarity property of all sampling consistent phylogenetic models except one

Neutral macroevolutionary models, such as the Yule model, give rise to a probability distribution on the set of discrete rooted binary trees over a given leaf set. Such models can provide a signal as to the approximate location of the root when only the unrooted phylogenetic tree is known, and this signal becomes relatively more significant as the number of leaves grows. In this short note, we show that among models that treat all taxa equally, and are sampling consistent (i.e. the distribution on trees is not affected by taxa yet to be included), all such models, except one (the so-called PDA model), convey some information as to the location of the ancestral root in an unrooted tree.     

Support vector machines for prediction of protein subcellular location by incorporating quasi-sequence-order effect.

Support Vector Machine (SVM), which is one class of learning machines, was applied to predict the subcellular location of proteins by incorporating the quasi-sequence-order effect (Chou [2000] Biochem. Biophys. Res. Commun. 278:477-483). In this study, the proteins are classified into the following 12 groups: (1) chloroplast, (2) cytoplasm, (3) cytoskeleton, (4) endoplasmic reticulum, (5) extracellular, (6) Golgi apparatus, (7) lysosome, (8) mitochondria, (9) nucleus, (10) peroxisome, (11) plasma membrane, and (12) vacuole, which account for most organelles and subcellular compartments in an animal or plant cell. Examinations for self-consistency and jackknife testing of the SVMs method were conducted for three sets consisting of 1,911, 2,044, and 2,191 proteins. The correct rates for self-consistency and the jackknife test values achieved with these protein sets were 94 and 83% for 1,911 proteins, 92 and 78% for 2,044 proteins, and 89 and 75% for 2,191 proteins, respectively. Furthermore, tests for correct prediction rates were undertaken with three independent testing datasets containing 2,148 proteins, 2,417 proteins, and 2,494 proteins producing values of 84, 77, and 74%, respectively.     

Toward protein engineering for phytoremediation: possibilities and challenges

The combination of rational protein engineering and directed evolution techniques allow for the redesign of enzymes with tailored properties for use in environmental remediation. This review summarizes current molecular methods for either altering or improving protein function and highlights examples of how these methods can address bioremediation problems. Although much of the protein engineering applied to environmental clean-up employs microbial systems, there is great potential for and significant challenges to translating these approaches to plant systems for phytoremediation purposes. Protein engineering technologies combined with genomic information and metabolic engineering strategies hold promise for the design of plants and microbes to remediate organic and inorganic pollutants.     

Toward accurate prediction of pKa values for internal protein residues: the importance of conformational relaxation and desolvation energy

Proton uptake or release controls many important biological processes, such as energy transduction, virus replication, and catalysis. Accurate pK(a) prediction informs about proton pathways, thereby revealing detailed acid-base mechanisms. Physics-based methods in the framework of molecular dynamics simulations not only offer pK(a) predictions but also inform about the physical origins of pK(a) shifts and provide details of ionization-induced conformational relaxation and large-scale transitions. One such method is the recently developed continuous constant pH molecular dynamics (CPHMD) method, which has been shown to be an accurate and robust pK(a) prediction tool for naturally occurring titratable residues. To further examine the accuracy and limitations of CPHMD, we blindly predicted the pK(a) values for 87 titratable residues introduced in various hydrophobic regions of staphylococcal nuclease and variants. The predictions gave a root-mean-square deviation of 1.69 pK units from experiment, and there were only two pK(a)'s with errors greater than 3.5 pK units. Analysis of the conformational fluctuation of titrating side-chains in the context of the errors of calculated pK(a) values indicate that explicit treatment of conformational flexibility and the associated dielectric relaxation gives CPHMD a distinct advantage. Analysis of the sources of errors suggests that more accurate pK(a) predictions can be obtained for the most deeply buried residues by improving the accuracy in calculating desolvation energies. Furthermore, it is found that the generalized Born implicit-solvent model underlying the current CPHMD implementation slightly distorts the local conformational environment such that the inclusion of an explicit-solvent representation may offer improvement of accuracy.     

Toward prediction of functional protein pockets using blind docking and pocket search algorithms

Location of functional binding pockets of bioactive ligands on protein molecules is essential in structural genomics and drug design projects. If the experimental determination of ligand-protein complex structures is complicated, blind docking (BD) and pocket search (PS) calculations can help in the prediction of atomic resolution binding mode and the location of the pocket of a ligand on the entire protein surface. Whereas the number of successful predictions by these methods is increasing even for the complicated cases of exosites or allosteric binding sites, their reliability has not been fully established. For a critical assessment of reliability, we use a set of ligand-protein complexes, which were found to be problematic in previous studies. The robustness of BD and PS methods is addressed in terms of success of the selection of truly functional pockets from among the many putative ones identified on the surfaces of ligand-bound and ligand-free (holo and apo) protein forms. Issues related to BD such as effect of hydration, existence of multiple pockets, and competition of subsidiary ligands are considered. Practical cases of PS are discussed, categorized and strategies are recommended for handling the different situations. PS can be used in conjunction with BD, as we find that a consensus approach combining the techniques improves predictive power.     

Hum-mPLoc: an ensemble classifier for large-scale human protein subcellular location prediction by incorporating samples with multiple sites

Proteins may simultaneously exist at, or move between, two or more different subcellular locations. Proteins with multiple locations or dynamic feature of this kind are particularly interesting because they may have some very special biological functions intriguing to investigators in both basic research and drug discovery. For instance, among the 6408 human protein entries that have experimentally observed subcellular location annotations in the Swiss-Prot database (version 50.7, released 19-Sept-2006), 973 ( approximately 15%) have multiple location sites. The number of total human protein entries (except those annotated with "fragment" or those with less than 50 amino acids) in the same database is 14,370, meaning a gap of (14,370-6408)=7962 entries for which no knowledge is available about their subcellular locations. Although one can use the computational approach to predict the desired information for the gap, so far all the existing methods for predicting human protein subcellular localization are limited in the case of single location site only. To overcome such a barrier, a new ensemble classifier, named Hum-mPLoc, was developed that can be used to deal with the case of multiple location sites as well. Hum-mPLoc is freely accessible to the public as a web server at http://202.120.37.186/bioinf/hum-multi. Meanwhile, for the convenience of people working in the relevant areas, Hum-mPLoc has been used to identify all human protein entries in the Swiss-Prot database that do not have subcellular location annotations or are annotated as being uncertain. The large-scale results thus obtained have been deposited in a downloadable file prepared with Microsoft Excel and named "Tab_Hum-mPLoc.xls". This file is available at the same website and will be updated twice a year to include new entries of human proteins and reflect the continuous development of Hum-mPLoc.     

Contact prediction in protein modeling: Scoring,folding and refinement of coarse-grained models

Background  

Several different methods for contact prediction succeeded within the Sixth Critical Assessment of Techniques for Protein Structure Prediction (CASP6). The most relevant were non-local contact predictions for targets from the most difficult categories: fold recognition-analogy and new fold. Such contacts could provide valuable structural information in case a template structure cannot be found in the PDB.     

Comparison of Arrhenius-type and Bêlehrádek-type models for prediction of bacterial growth in foods

D.A. RATKOWSKY, T. ROSS, T.A. WCMEEKIN AND J. OLLEY. 1991. The development of Arrhenius-type ('Schoolfield') and Bêlehrádek-type (square root) models that describe microbial growth rates is briefly described. Both types of model have been advocated for use in predictive microbiology. On the basis of published data sets for the growth of bacteria, the consequences of mathematical transformation of data and the use of invalid stochastic assumptions upon model predictions are demonstrated. Mean square error is shown to be an inappropriate criterion by which to compare the performance of predictive models. The data show that bacterial growth responses such as generation time and lag time become more variable as their mean magnitude increases. The practical consequences of such variability for predictive microbiology are discussed.