HMMpTM: Improving transmembrane protein topology prediction using phosphorylation and glycosylation site prediction

During the last two decades a large number of computational methods have been developed for predicting transmembrane protein topology. Current predictors rely on topogenic signals in the protein sequence, such as the distribution of positively charged residues in extra-membrane loops and the existence of N-terminal signals. However, phosphorylation and glycosylation are post-translational modifications (PTMs) that occur in a compartment-specific manner and therefore the presence of a phosphorylation or glycosylation site in a transmembrane protein provides topological information. We examine the combination of phosphorylation and glycosylation site prediction with transmembrane protein topology prediction. We report the development of a Hidden Markov Model based method, capable of predicting the topology of transmembrane proteins and the existence of kinase specific phosphorylation and N/O-linked glycosylation sites along the protein sequence. Our method integrates a novel feature in transmembrane protein topology prediction, which results in improved performance for topology prediction and reliable prediction of phosphorylation and glycosylation sites. The method is freely available at http://bioinformatics.biol.uoa.gr/HMMpTM.     

Meta-prediction of phosphorylation sites with weighted voting and restricted grid search parameter selection

Meta-predictors make predictions by organizing and processing the predictions produced by several other predictors in a defined problem domain. A proficient meta-predictor not only offers better predicting performance than the individual predictors from which it is constructed, but it also relieves experimentally researchers from making difficult judgments when faced with conflicting results made by multiple prediction programs. As increasing numbers of predicting programs are being developed in a large number of fields of life sciences, there is an urgent need for effective meta-prediction strategies to be investigated. We compiled four unbiased phosphorylation site datasets, each for one of the four major serine/threonine (S/T) protein kinase families-CDK, CK2, PKA and PKC. Using these datasets, we examined several meta-predicting strategies with 15 phosphorylation site predictors from six predicting programs: GPS, KinasePhos, NetPhosK, PPSP, PredPhospho and Scansite. Meta-predictors constructed with a generalized weighted voting meta-predicting strategy with parameters determined by restricted grid search possess the best performance, exceeding that of all individual predictors in predicting phosphorylation sites of all four kinase families. Our results demonstrate a useful decision-making tool for analysing the predictions of the various S/T phosphorylation site predictors. An implementation of these meta-predictors is available on the web at: http://MetaPred.umn.edu/MetaPredPS/.     

SiteSeek: Post-translational modification analysis using adaptive locality-effective kernel methods and new profiles

Background  

Post-translational modifications have a substantial influence on the structure and functions of protein. Post-translational phosphorylation is one of the most common modification that occur in intracellular proteins. Accurate prediction of protein phosphorylation sites is of great importance for the understanding of diverse cellular signalling processes in both the human body and in animals. In this study, we propose a new machine learning based protein phosphorylation site predictor, SiteSeek. SiteSeek is trained using a novel compact evolutionary and hydrophobicity profile to detect possible protein phosphorylation sites for a target sequence. The newly proposed method proves to be more accurate and exhibits a much stable predictive performance than currently existing phosphorylation site predictors.     

NetPhosBac – A predictor for Ser/Thr phosphorylation sites in bacterial proteins

There is ample evidence for the involvement of protein phosphorylation on serine/threonine/tyrosine in bacterial signaling and regulation, but very few exact phosphorylation sites have been experimentally determined. Recently, gel‐free high accuracy MS studies reported over 150 phosphorylation sites in two bacterial model organisms Bacillus subtilis and Escherichia coli. Interestingly, the analysis of these phosphorylation sites revealed that most of them are not characteristic for eukaryotic‐type protein kinases, which explains the poor performance of eukaryotic data‐trained phosphorylation predictors on bacterial systems. We used these large bacterial datasets and neural network algorithms to create the first bacteria‐specific protein phosphorylation predictor: NetPhosBac. With respect to predicting bacterial phosphorylation sites, NetPhosBac significantly outperformed all benchmark predictors. Moreover, NetPhosBac predictions of phosphorylation sites in E. coli proteins were experimentally verified on protein and site‐specific levels. In conclusion, NetPhosBac clearly illustrates the advantage of taxa‐specific predictors and we hope it will provide a useful asset to the microbiological community.     

Prediction of kinase-specific phosphorylation sites with sequence features by a log-odds ratio approach

Protein phosphorylation plays important roles in a variety of cellular processes. Detecting possible phosphorylation sites and their corresponding protein kinases is crucial for studying the function of many proteins. This article presents a new prediction system, called PhoScan, to predict phosphorylation sites in a kinase-family-specific way. Common phosphorylation features and kinase-specific features are extracted from substrate sequences of different protein kinases based on the analysis of published experiments, and a scoring system is developed for evaluating the possibility that a peptide can be phosphorylated by the protein kinase at the specific site in its sequence context. PhoScan can achieve a specificity of above 90% with sensitivity around 90% at kinase-family level on the data experimented. The system is applied on a set of human proteins collected from Swiss-Prot and sets of putative phosphorylation sites are predicted for protein kinase A, cyclin-dependent kinase, and casein kinase 2 families. PhoScan is available at http://bioinfo.au.tsinghua.edu.cn/phoscan/.     

A novel sequence-based method for phosphorylation site prediction with feature selection and analysis

Phosphorylation is one of the most important post-translational modifications, and the identification of protein phosphorylation sites is particularly important for studying disease diagnosis. However, experimental detection of phosphorylation sites is labor intensive. It would be beneficial if computational methods are available to provide an extra reference for the phosphorylation sites. Here we developed a novel sequence-based method for serine, threonine, and tyrosine phosphorylation site prediction. Nearest Neighbor algorithm was employed as the prediction engine. The peptides around the phosphorylation sites with a fixed length of thirteen amino acid residues were extracted via a sliding window along the protein chains concerned. Each of such peptides was coded into a vector with 6,072 features, derived from Amino Acid Index (AAIndex) database, for the classification/detection. Incremental Feature Selection, a feature selection algorithm based on the Maximum Relevancy Minimum Redundancy (mRMR) method was used to select a compact feature set for a further improvement of the classification performance. Three predictors were established for identifying the three types of phosphorylation sites, achieving the overall accuracies of 66.64%, 66.11%% and 66.69%, respectively. These rates were obtained by rigorous jackknife cross-validation tests.     

Prediction of protein kinase-specific phosphorylation sites in hierarchical structure using functional information and random forest

Reversible protein phosphorylation is one of the most important post-translational modifications, which regulates various biological cellular processes. Identification of the kinase-specific phosphorylation sites is helpful for understanding the phosphorylation mechanism and regulation processes. Although a number of computational approaches have been developed, currently few studies are concerned about hierarchical structures of kinases, and most of the existing tools use only local sequence information to construct predictive models. In this work, we conduct a systematic and hierarchy-specific investigation of protein phosphorylation site prediction in which protein kinases are clustered into hierarchical structures with four levels including kinase, subfamily, family and group. To enhance phosphorylation site prediction at all hierarchical levels, functional information of proteins, including gene ontology (GO) and protein–protein interaction (PPI), is adopted in addition to primary sequence to construct prediction models based on random forest. Analysis of selected GO and PPI features shows that functional information is critical in determining protein phosphorylation sites for every hierarchical level. Furthermore, the prediction results of Phospho.ELM and additional testing dataset demonstrate that the proposed method remarkably outperforms existing phosphorylation prediction methods at all hierarchical levels. The proposed method is freely available at http://bioinformatics.ustc.edu.cn/phos_pred/.     

Prediction of functional phosphorylation sites by incorporating evolutionary information

Protein phosphorylation is a ubiquitous protein post-translational modification, which plays an important role in cellular signaling systems underlying various physiological and pathological processes. Current in silico methods mainly focused on the prediction of phosphorylation sites, but rare methods considered whether a phosphorylation site is functional or not. Since functional phosphorylation sites are more valuable for further experimental research and a proportion of phosphorylation sites have no direct functional effects, the prediction of functional phosphorylation sites is quite necessary for this research area. Previous studies have shown that functional phosphorylation sites are more conserved than non-functional phosphorylation sites in evolution. Thus, in our method, we developed a web server by integrating existing phosphorylation site prediction methods, as well as both absolute and relative evolutionary conservation scores to predict the most likely functional phosphorylation sites. Using our method, we predicted the most likely functional sites of the human, rat and mouse proteomes and built a database for the predicted sites. By the analysis of overall prediction results, we demonstrated that protein phosphorylation plays an important role in all the enriched KEGG pathways. By the analysis of protein-specific prediction results, we demonstrated the usefulness of our method for individual protein studies. Our method would help to characterize the most likely functional phosphorylation sites for further studies in this research area.     

Highly specific prediction of phosphorylation sites in proteins

SUMMARY: The prediction of significant short functional protein sequences has inherent problems. In predicting phosphorylation sites, problems came from the shortness of phosphorylation sites, the difficulties in maintaining many different predefined models of binding sites, and the difficulties of obtaining highly sensitive predictions and of obtaining predictions with a constant sensitivity and specificity. The algorithm presented in this paper overcomes these problems. The proposed algorithm PHOSITE is based on the case-based sequence analysis. This enables the prediction of phosphorylation sites with constant specificity and sensitivity. Furthermore, this method leads not only to the prediction of phosphorylation sites in general but also predicts the most probable type of kinase involved. AVAILABILITY: The tool PHOSITE implementing the presented method can be evaluated under the website http://www.phosite.com.     

Prediction of carbohydrate binding sites on protein surfaces with 3-dimensional probability density distributions of interacting atoms

Non-covalent protein-carbohydrate interactions mediate molecular targeting in many biological processes. Prediction of non-covalent carbohydrate binding sites on protein surfaces not only provides insights into the functions of the query proteins; information on key carbohydrate-binding residues could suggest site-directed mutagenesis experiments, design therapeutics targeting carbohydrate-binding proteins, and provide guidance in engineering protein-carbohydrate interactions. In this work, we show that non-covalent carbohydrate binding sites on protein surfaces can be predicted with relatively high accuracy when the query protein structures are known. The prediction capabilities were based on a novel encoding scheme of the three-dimensional probability density maps describing the distributions of 36 non-covalent interacting atom types around protein surfaces. One machine learning model was trained for each of the 30 protein atom types. The machine learning algorithms predicted tentative carbohydrate binding sites on query proteins by recognizing the characteristic interacting atom distribution patterns specific for carbohydrate binding sites from known protein structures. The prediction results for all protein atom types were integrated into surface patches as tentative carbohydrate binding sites based on normalized prediction confidence level. The prediction capabilities of the predictors were benchmarked by a 10-fold cross validation on 497 non-redundant proteins with known carbohydrate binding sites. The predictors were further tested on an independent test set with 108 proteins. The residue-based Matthews correlation coefficient (MCC) for the independent test was 0.45, with prediction precision and sensitivity (or recall) of 0.45 and 0.49 respectively. In addition, 111 unbound carbohydrate-binding protein structures for which the structures were determined in the absence of the carbohydrate ligands were predicted with the trained predictors. The overall prediction MCC was 0.49. Independent tests on anti-carbohydrate antibodies showed that the carbohydrate antigen binding sites were predicted with comparable accuracy. These results demonstrate that the predictors are among the best in carbohydrate binding site predictions to date.     

Prediction of PK-specific phosphorylation site based on information entropy

Phosphorylation is a crucial way to control the activity of proteins in many eukaryotic organisms in vivo. Experimental methods to determine phosphorylation sites in substrates are usually restricted by the in vitro condition of enzymes and very intensive in time and labor. Although some in silico methods and web servers have been introduced for automatic detection of phosphorylation sites, sophisticated methods are still in urgent demand to further improve prediction performances. Protein primary se-quences can help predict phosphorylation sites catalyzed by different protein kinase and most com-putational approaches use a short local peptide to make prediction. However, the useful information may be lost if only the conservative residues that are not close to the phosphorylation site are consid-ered in prediction, which would hamper the prediction results. A novel prediction method named IEPP (Information-Entropy based Phosphorylation Prediction) is presented in this paper for automatic de-tection of potential phosphorylation sites. In prediction, the sites around the phosphorylation sites are selected or excluded by their entropy values. The algorithm was compared with other methods such as GSP and PPSP on the ABL, MAPK and PKA PK families. The superior prediction accuracies were ob-tained in various measurements such as sensitivity (Sn) and specificity (Sp). Furthermore, compared with some online prediction web servers on the new discovered phosphorylation sites, IEPP also yielded the best performance. IEPP is another useful computational resource for identification of PK-specific phosphorylation sites and it also has the advantages of simpleness, efficiency and con-venience.     

Prediction of PK-specific phosphorylation site based on information entropy

Phosphorylation is a crucial way to control the activity of proteins in many eukaryotic organisms in vivo. Experimental methods to determine phosphorylation sites in substrates are usually restricted by the in vitro condition of enzymes and very intensive in time and labor. Although some in silico methods and web servers have been introduced for automatic detection of phosphorylation sites, sophisticated methods are still in urgent demand to further improve prediction performances. Protein primary sequences can help predict phosphorylation sites catalyzed by different protein kinase and most computational approaches use a short local peptide to make prediction. However, the useful information may be lost if only the conservative residues that are not close to the phosphorylation site are considered in prediction, which would hamper the prediction results. A novel prediction method named IEPP (Information-Entropy based Phosphorylation Prediction) is presented in this paper for automatic detection of potential phosphorylation sites. In prediction, the sites around the phosphorylation sites are selected or excluded by their entropy values. The algorithm was compared with other methods such as GSP and PPSP on the ABL, MAPK and PKA PK families. The superior prediction accuracies were obtained in various measurements such as sensitivity (Sn) and specificity (Sp). Furthermore, compared with some online prediction web servers on the new discovered phosphorylation sites, IEPP also yielded the best performance. IEPP is another useful computational resource for identification of PK-specific phosphorylation sites and it also has the advantages of simpleness, efficiency and convenience.     

A novel mechanism of enzymic ester hydrolysis. Inversion of configuration and carbon-oxygen bond cleavage by secondary alkylsulphohydrolases from detergent-degrading micro-organisms.

Ligand-binding studies with labelled triethyltin on yeast mitochondrial membranes showed the presence of high-affinity sites (KD = 0.6 micronM; 1.2 +/- 0.3 nmol/mg of protein) and low-affinity sites (KD less than 45 micronM; 70 +/- 20 nmol/mg of protein). The dissociation constant of the high-affinity site is in good agreement with the concentration of triethyltin required for inhibition of mitochondrial ATPase (adenosine triphosphatase) and oxidative phosphorylation. The high-affinity site is not competed for by oligomycin or venturicidin, indicating that triethyltin reacts at a different site from these inhibitors of oxidative phosphorylation. Fractionation of the mitochondrial membrane shows a specific association of the high-affinity sites with the ATP synthase complex. During purification of ATP synthase (oligomycin-sensitive ATPase) there is a 5-6-fold purification of oligomycin- and triethyltin-sensitive ATPase activity concomitant with a 7-9-fold increase in high-affinity triethyltin-binding sites. The purified yeast oligomycin-sensitive ATPase complex contains approximately six binding sites for triethyltin/mol of enzyme complex. It is concluded that specific triethyltin-binding sites are components of the ATP synthase complex, which accounts for the specific inhibition of ATPase and oxidative phosphorylation by triethyltin.     

GPS: a novel group-based phosphorylation predicting and scoring method

Protein phosphorylation is an important reversible post-translational modification of proteins, and it orchestrates a variety of cellular processes. Experimental identification of phosphorylation site is labor-intensive and often limited by the availability and optimization of enzymatic reaction. In silico prediction may facilitate the identification of potential phosphorylation sites with ease. Here we present a novel computational method named GPS: group-based phosphorylation site predicting and scoring platform. If two polypeptides differ by only two consecutive amino acids, in particular when the two different amino acids are a conserved pair, e.g., isoleucine (I) and valine (V), or serine (S) and threonine (T), we view these two polypeptides bearing similar 3D structures and biochemical properties. Based on this rationale, we formulated GPS that carries greater computational power with superior performance compared to two existing phosphorylation sites prediction systems, ScanSite 2.0 and PredPhospho. With database in public domain, GPS can predict substrate phosphorylation sites from 52 different protein kinase (PK) families while ScanSite 2.0 and PredPhospho offer at most 30 PK families. Using PKA as a model enzyme, we first compared prediction profiles from the GPS method with those from ScanSite 2.0 and PredPhospho. In addition, we chose an essential mitotic kinase Aurora-B as a model enzyme since ScanSite 2.0 and PredPhospho offer no prediction. However, GPS offers satisfactory sensitivity (94.44%) and specificity (97.14%). Finally, the accuracy of phosphorylation on MCAK predicted by GPS was validated by experimentation, in which six out of seven predicted potential phosphorylation sites on MCAK (Q91636) were experimentally verified. Taken together, we have generated a novel method to predict phosphorylation sites, which offers greater precision and computing power over ScanSite 2.0 and PredPhospho.     

Proteome-wide prediction of PKA phosphorylation sites in eukaryotic kingdom

Protein phosphorylation is one of the most essential post-translational modifications (PTMs), and orchestrates a variety of cellular functions and processes. Besides experimental studies, numerous computational predictors implemented in various algorithms have been developed for phosphorylation sites prediction. However, large-scale predictions of kinase-specific phosphorylation sites have not been successfully pursued and remained to be a great challenge. In this work, we raised a “kiss farewell” model and conducted a high-throughput prediction of cAMP-dependent kinase (PKA) phosphorylation sites. Since a protein kinase (PK) should at least “kiss” its substrates and then run away, we proposed a PKA-binding protein to be a potential PKA substrate if at least one PKA site was predicted. To improve the prediction specificity, we reduced false positive rate (FPR) less than 1% when the cut-off value was set as 4. Successfully, we predicted 1387, 630, 568 and 912 potential PKA sites from 410, 217, 173 and 260 PKA-interacting proteins in Saccharomyces cerevisiae, Caenorhabditis elegans, Drosophila melanogaster and Homo sapiens, respectively. Most of these potential phosphorylation sites remained to be experimentally verified. In addition, we detected two sites in one of PKA regulatory subunits to be conserved in eukaryotes as potentially ancient regulatory signals. Our prediction results provide an excellent resource for delineating PKA-mediated signaling pathways and their system integration underlying cellular dynamics and plasticity.     

Prediction of ubiquitination sites by using the composition of k-spaced amino acid pairs

As one of the most important reversible protein post-translation modifications, ubiquitination has been reported to be involved in lots of biological processes and closely implicated with various diseases. To fully decipher the molecular mechanisms of ubiquitination-related biological processes, an initial but crucial step is the recognition of ubiquitylated substrates and the corresponding ubiquitination sites. Here, a new bioinformatics tool named CKSAAP_UbSite was developed to predict ubiquitination sites from protein sequences. With the assistance of Support Vector Machine (SVM), the highlight of CKSAAP_UbSite is to employ the composition of k-spaced amino acid pairs surrounding a query site (i.e. any lysine in a query sequence) as input. When trained and tested in the dataset of yeast ubiquitination sites (Radivojac et al, Proteins, 2010, 78: 365-380), a 100-fold cross-validation on a 1∶1 ratio of positive and negative samples revealed that the accuracy and MCC of CKSAAP_UbSite reached 73.40% and 0.4694, respectively. The proposed CKSAAP_UbSite has also been intensively benchmarked to exhibit better performance than some existing predictors, suggesting that it can be served as a useful tool to the community. Currently, CKSAAP_UbSite is freely accessible at http://protein.cau.edu.cn/cksaap_ubsite/. Moreover, we also found that the sequence patterns around ubiquitination sites are not conserved across different species. To ensure a reasonable prediction performance, the application of the current CKSAAP_UbSite should be limited to the proteome of yeast.     

Prediction of phosphorylation sites using SVMs

MOTIVATION: Phosphorylation is involved in diverse signal transduction pathways. By predicting phosphorylation sites and their kinases from primary protein sequences, we can obtain much valuable information that can form the basis for further research. Using support vector machines, we attempted to predict phosphorylation sites and the type of kinase that acts at each site. RESULTS: Our prediction system was limited to phosphorylation sites catalyzed by four protein kinase families and four protein kinase groups. The accuracy of the predictions ranged from 83 to 95% at the kinase family level, and 76-91% at the kinase group level. The prediction system used-PredPhospho-can be applied to the functional study of proteins, and can help predict the changes in phosphorylation sites caused by amino acid variations at intra- and interspecies levels.