Using support vector machine combined with auto covariance to predict protein-protein interactions from protein sequences

Compared to the available protein sequences of different organisms, the number of revealed protein-protein interactions (PPIs) is still very limited. So many computational methods have been developed to facilitate the identification of novel PPIs. However, the methods only using the information of protein sequences are more universal than those that depend on some additional information or predictions about the proteins. In this article, a sequence-based method is proposed by combining a new feature representation using auto covariance (AC) and support vector machine (SVM). AC accounts for the interactions between residues a certain distance apart in the sequence, so this method adequately takes the neighbouring effect into account. When performed on the PPI data of yeast Saccharomyces cerevisiae, the method achieved a very promising prediction result. An independent data set of 11,474 yeast PPIs was used to evaluate this prediction model and the prediction accuracy is 88.09%. The performance of this method is superior to those of the existing sequence-based methods, so it can be a useful supplementary tool for future proteomics studies. The prediction software and all data sets used in this article are freely available at http://www.scucic.cn/Predict_PPI/index.htm.     

Adaptive compressive learning for prediction of protein-protein interactions from primary sequence

Protein-protein interactions (PPIs) play an important role in biological processes. Although much effort has been devoted to the identification of novel PPIs by integrating experimental biological knowledge, there are still many difficulties because of lacking enough protein structural and functional information. It is highly desired to develop methods based only on amino acid sequences for predicting PPIs. However, sequence-based predictors are often struggling with the high-dimensionality causing over-fitting and high computational complexity problems, as well as the redundancy of sequential feature vectors. In this paper, a novel computational approach based on compressed sensing theory is proposed to predict yeast Saccharomyces cerevisiae PPIs from primary sequence and has achieved promising results. The key advantage of the proposed compressed sensing algorithm is that it can compress the original high-dimensional protein sequential feature vector into a much lower but more condensed space taking the sparsity property of the original signal into account. What makes compressed sensing much more attractive in protein sequence analysis is its compressed signal can be reconstructed from far fewer measurements than what is usually considered necessary in traditional Nyquist sampling theory. Experimental results demonstrate that proposed compressed sensing method is powerful for analyzing noisy biological data and reducing redundancy in feature vectors. The proposed method represents a new strategy of dealing with high-dimensional protein discrete model and has great potentiality to be extended to deal with many other complicated biological systems.     

Inferring domain-domain interactions from protein-protein interactions in the complex network conformation

Background

As protein domains are functional and structural units of proteins, a large proportion of protein-protein interactions (PPIs) are achieved by domain-domain interactions (DDIs), many computational efforts have been made to identify DDIs from experimental PPIs since high throughput technologies have produced a large number of PPIs for different species. These methods can be separated into two categories: deterministic and probabilistic. In deterministic methods, parsimony assumption has been utilized. Parsimony principle has been widely used in computational biology as the evolution of the nature is considered as a continuous optimization process. In the context of identifying DDIs, parsimony methods try to find a minimal set of DDIs that can explain the observed PPIs. This category of methods are promising since they can be formulated and solved easily. Besides, researches have shown that they can detect specific DDIs, which is often hard for many probabilistic methods. We notice that existing methods just view PPI networks as simply assembled by single interactions, but there is now ample evidence that PPI networks should be considered in a global (systematic) point of view for it exhibits general properties of complex networks, such as 'scale-free' and 'small-world'.

Results

In this work, we integrate this global point of view into the parsimony-based model. Particularly, prior knowledge is extracted from these global properties by plausible reasoning and then taken as input. We investigate the role of the added information extensively through numerical experiments. Results show that the proposed method has improved performance, which confirms the biological meanings of the extracted prior knowledge.

Conclusions

This work provides us some clues for using these properties of complex networks in computational models and to some extent reveals the biological meanings underlying these general network properties.

     

DomPep--a general method for predicting modular domain-mediated protein-protein interactions

Protein-protein interactions (PPIs) are frequently mediated by the binding of a modular domain in one protein to a short, linear peptide motif in its partner. The advent of proteomic methods such as peptide and protein arrays has led to the accumulation of a wealth of interaction data for modular interaction domains. Although several computational programs have been developed to predict modular domain-mediated PPI events, they are often restricted to a given domain type. We describe DomPep, a method that can potentially be used to predict PPIs mediated by any modular domains. DomPep combines proteomic data with sequence information to achieve high accuracy and high coverage in PPI prediction. Proteomic binding data were employed to determine a simple yet novel parameter Ligand-Binding Similarity which, in turn, is used to calibrate Domain Sequence Identity and Position-Weighted-Matrix distance, two parameters that are used in constructing prediction models. Moreover, DomPep can be used to predict PPIs for both domains with experimental binding data and those without. Using the PDZ and SH2 domain families as test cases, we show that DomPep can predict PPIs with accuracies superior to existing methods. To evaluate DomPep as a discovery tool, we deployed DomPep to identify interactions mediated by three human PDZ domains. Subsequent in-solution binding assays validated the high accuracy of DomPep in predicting authentic PPIs at the proteome scale. Because DomPep makes use of only interaction data and the primary sequence of a domain, it can be readily expanded to include other types of modular domains.     

Computational prediction of host-pathogen protein-protein interactions

MOTIVATION: Infectious diseases such as malaria result in millions of deaths each year. An important aspect of any host-pathogen system is the mechanism by which a pathogen can infect its host. One method of infection is via protein-protein interactions (PPIs) where pathogen proteins target host proteins. Developing computational methods that identify which PPIs enable a pathogen to infect a host has great implications in identifying potential targets for therapeutics. RESULTS: We present a method that integrates known intra-species PPIs with protein-domain profiles to predict PPIs between host and pathogen proteins. Given a set of intra-species PPIs, we identify the functional domains in each of the interacting proteins. For every pair of functional domains, we use Bayesian statistics to assess the probability that two proteins with that pair of domains will interact. We apply our method to the Homo sapiens-Plasmodium falciparum host-pathogen system. Our system predicts 516 PPIs between proteins from these two organisms. We show that pairs of human proteins we predict to interact with the same Plasmodium protein are close to each other in the human PPI network and that Plasmodium pairs predicted to interact with same human protein are co-expressed in DNA microarray datasets measured during various stages of the Plasmodium life cycle. Finally, we identify functionally enriched sub-networks spanned by the predicted interactions and discuss the plausibility of our predictions. AVAILABILITY: Supplementary data are available at http://staff.vbi.vt.edu/dyermd/publications/dyer2007a.html. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.     

Predicting protein-protein interactions by combing various sequence- derived features into the general form of Chou's Pseudo amino acid composition

Knowledge of protein-protein interactions (PPIs) plays an important role in constructing protein interaction networks and understanding the general machineries of biological systems. In this study, a new method is proposed to predict PPIs using a comprehensive set of 930 features based only on sequence information, these features measure the interactions between residues a certain distant apart in the protein sequences from different aspects. To achieve better performance, the principal component analysis (PCA) is first employed to obtain an optimized feature subset. Then, the resulting 67-dimensional feature vectors are fed to Support Vector Machine (SVM). Experimental results on Drosophila melanogaster and Helicobater pylori datasets show that our method is very promising to predict PPIs and may at least be a useful supplement tool to existing methods.     

Detecting protein complexes based on a combination of topological and biological properties in protein-protein interaction network

Protein complexes are known to play a major role in controlling cellular activity in a living being. Identifying complexes from raw protein protein interactions (PPIs) is an important area of research. Earlier work has been limited mostly to yeast. Such protein complex identification methods, when applied to large human PPIs often give poor performance. We introduce a novel method called CSC to detect protein complexes. The method is evaluated in terms of positive predictive value, sensitivity and accuracy using the datasets of the model organism, yeast and humans. CSC outperforms several other competing algorithms for both organisms. Further, we present a framework to establish the usefulness of CSC in analyzing the influence of a given disease gene in a complex topologically as well as biologically considering eight major association factors.     

A mouse protein interactome through combined literature mining with multiple sources of interaction evidence

Protein–protein interactions (PPIs) play crucial roles in a number of biological processes. Recently, protein interaction networks (PINs) for several model organisms and humans have been generated, but few large-scale researches for mice have ever been made neither experimentally nor computationally. In the work, we undertook an effort to map a mouse PIN, in which protein interactions are hidden in enormous amount of biomedical literatures. Following a co-occurrence-based text-mining approach, a probabilistic model—naïve Bayesian was used to filter false-positive interactions by integrating heterogeneous kinds of evidence from genomic and proteomic datasets. A support vector machine algorithm was further used to choose protein pairs with physical interactions. By comparing with the currently available PPI datasets from several model organisms and humans, it showed that the derived mouse PINs have similar topological properties at the global level, but a high local divergence. The mouse protein interaction dataset is stored in the Mouse protein–protein interaction DataBase (MppDB) that is useful source of information for system-level understanding of gene function and biological processes in mammals. Access to the MppDB database is public available at http://bio.scu.edu.cn/mppi.     

A computational model for predicting protein interactions based on multidomain collaboration

Recently, several domain-based computational models for predicting protein-protein interactions (PPIs) have been proposed. The conventional methods usually infer domain or domain combination (DC) interactions from already known interacting sets of proteins, and then predict PPIs using the information. However, the majority of these models often have limitations in providing detailed information on which domain pair (single domain interaction) or DC pair (multidomain interaction) will actually interact for the predicted protein interaction. Therefore, a more comprehensive and concrete computational model for the prediction of PPIs is needed. We developed a computational model to predict PPIs using the information of intraprotein domain cohesion and interprotein DC coupling interaction. A method of identifying the primary interacting DC pair was also incorporated into the model in order to infer actual participants in a predicted interaction. Our method made an apparent improvement in the PPI prediction accuracy, and the primary interacting DC pair identification was valid specifically in predicting multidomain protein interactions. In this paper, we demonstrate that 1) the intraprotein domain cohesion is meaningful in improving the accuracy of domain-based PPI prediction, 2) a prediction model incorporating the intradomain cohesion enables us to identify the primary interacting DC pair, and 3) a hybrid approach using the intra/interdomain interaction information can lead to a more accurate prediction.     

Bayesian inference for genomic data integration reduces misclassification rate in predicting protein-protein interactions

Protein-protein interactions (PPIs) are essential to most fundamental cellular processes. There has been increasing interest in reconstructing PPIs networks. However, several critical difficulties exist in obtaining reliable predictions. Noticeably, false positive rates can be as high as >80%. Error correction from each generating source can be both time-consuming and inefficient due to the difficulty of covering the errors from multiple levels of data processing procedures within a single test. We propose a novel Bayesian integration method, deemed nonparametric Bayes ensemble learning (NBEL), to lower the misclassification rate (both false positives and negatives) through automatically up-weighting data sources that are most informative, while down-weighting less informative and biased sources. Extensive studies indicate that NBEL is significantly more robust than the classic naïve Bayes to unreliable, error-prone and contaminated data. On a large human data set our NBEL approach predicts many more PPIs than naïve Bayes. This suggests that previous studies may have large numbers of not only false positives but also false negatives. The validation on two human PPIs datasets having high quality supports our observations. Our experiments demonstrate that it is feasible to predict high-throughput PPIs computationally with substantially reduced false positives and false negatives. The ability of predicting large numbers of PPIs both reliably and automatically may inspire people to use computational approaches to correct data errors in general, and may speed up PPIs prediction with high quality. Such a reliable prediction may provide a solid platform to other studies such as protein functions prediction and roles of PPIs in disease susceptibility.     

Discovering disease-genes by topological features in human protein-protein interaction network

MOTIVATION: Mining the hereditary disease-genes from human genome is one of the most important tasks in bioinformatics research. A variety of sequence features and functional similarities between known human hereditary disease-genes and those not known to be involved in disease have been systematically examined and efficient classifiers have been constructed based on the identified common patterns. The availability of human genome-wide protein-protein interactions (PPIs) provides us with new opportunity for discovering hereditary disease-genes by topological features in PPIs network. RESULTS: This analysis reveals that the hereditary disease-genes ascertained from OMIM in the literature-curated (LC) PPIs network are characterized by a larger degree, tendency to interact with other disease-genes, more common neighbors and quick communication to each other whereas those properties could not be detected from the network identified from high-throughput yeast two-hybrid mapping approach (EXP) and predicted interactions (PDT) PPIs network. KNN classifier based on those features was created and on average gained overall prediction accuracy of 0.76 in cross-validation test. Then the classifier was applied to 5262 genes on human genome and predicted 178 novel disease-genes. Some of the predictions have been validated by biological experiments.     

Improving protein–protein interactions prediction accuracy using protein evolutionary information and relevance vector machine model

Predicting protein–protein interactions (PPIs) is a challenging task and essential to construct the protein interaction networks, which is important for facilitating our understanding of the mechanisms of biological systems. Although a number of high‐throughput technologies have been proposed to predict PPIs, there are unavoidable shortcomings, including high cost, time intensity, and inherently high false positive rates. For these reasons, many computational methods have been proposed for predicting PPIs. However, the problem is still far from being solved. In this article, we propose a novel computational method called RVM‐BiGP that combines the relevance vector machine (RVM) model and Bi‐gram Probabilities (BiGP) for PPIs detection from protein sequences. The major improvement includes (1) Protein sequences are represented using the Bi‐gram probabilities (BiGP) feature representation on a Position Specific Scoring Matrix (PSSM), in which the protein evolutionary information is contained; (2) For reducing the influence of noise, the Principal Component Analysis (PCA) method is used to reduce the dimension of BiGP vector; (3) The powerful and robust Relevance Vector Machine (RVM) algorithm is used for classification. Five‐fold cross‐validation experiments executed on yeast and Helicobacter pylori datasets, which achieved very high accuracies of 94.57 and 90.57%, respectively. Experimental results are significantly better than previous methods. To further evaluate the proposed method, we compare it with the state‐of‐the‐art support vector machine (SVM) classifier on the yeast dataset. The experimental results demonstrate that our RVM‐BiGP method is significantly better than the SVM‐based method. In addition, we achieved 97.15% accuracy on imbalance yeast dataset, which is higher than that of balance yeast dataset. The promising experimental results show the efficiency and robust of the proposed method, which can be an automatic decision support tool for future proteomics research. For facilitating extensive studies for future proteomics research, we developed a freely available web server called RVM‐BiGP‐PPIs in Hypertext Preprocessor (PHP) for predicting PPIs. The web server including source code and the datasets are available at http://219.219.62.123:8888/BiGP/ .     

Assembling a protein-protein interaction map of the SSU processome from existing datasets

Background

The small subunit (SSU) processome is a large ribonucleoprotein complex involved in small ribosomal subunit assembly. It consists of the U3 snoRNA and ∼72 proteins. While most of its components have been identified, the protein-protein interactions (PPIs) among them remain largely unknown, and thus the assembly, architecture and function of the SSU processome remains unclear.

Methodology

We queried PPI databases for SSU processome proteins to quantify the degree to which the three genome-wide high-throughput yeast two-hybrid (HT-Y2H) studies, the genome-wide protein fragment complementation assay (PCA) and the literature-curated (LC) datasets cover the SSU processome interactome.

Conclusions

We find that coverage of the SSU processome PPI network is remarkably sparse. Two of the three HT-Y2H studies each account for four and six PPIs between only six of the 72 proteins, while the third study accounts for as little as one PPI and two proteins. The PCA dataset has the highest coverage among the genome-wide studies with 27 PPIs between 25 proteins. The LC dataset was the most extensive, accounting for 34 proteins and 38 PPIs, many of which were validated by independent methods, thereby further increasing their reliability. When the collected data were merged, we found that at least 70% of the predicted PPIs have yet to be determined and 26 proteins (36%) have no known partners. Since the SSU processome is conserved in all Eukaryotes, we also queried HT-Y2H datasets from six additional model organisms, but only four orthologues and three previously known interologous interactions were found. This provides a starting point for further work on SSU processome assembly, and spotlights the need for a more complete genome-wide Y2H analysis.     

A versatile lentiviral expression system to identify mammalian protein-protein interactions

Protein-protein interactions (PPIs) are central to our understanding of protein function, biological processes and signaling pathways. Affinity purification coupled with mass spectrometry (AP-MS) is a powerful approach for detecting PPIs and protein complexes and relies on the purification of bait proteins using bait-specific binding reagents. These binding reagents may recognize bait proteins directly or affinity tags that are fused to bait proteins. A limitation of the latter approach is that expression of affinity tagged baits is largely constrained to engineered or unnatural cell lines, which results in the AP-MS identification of PPIs that may not accurately reflect those seen in nature. Therefore, generating cell lines stably expressing affinity tagged bait proteins in a broad range of cell types and cell lines is important for identifying PPIs that are dependent on different contexts. To facilitate the identification of PPIs across many mammalian cell types, we developed the mammalian affinity purification and lentiviral expression (MAPLE) system. MAPLE uses recombinant lentiviral technology to stably and efficiently express affinity tagged complementary DNA (cDNA) in mammalian cells, including cells that are difficult to transfect and non-dividing cells. The MAPLE vectors contain a versatile affinity (VA) tag for multi-step protein purification schemes and subcellular localization studies. In this methods article, we present a step-by-step overview of the MAPLE system workflow.     

Global investigation of protein-protein interactions in yeast Saccharomyces cerevisiae using re-occurring short polypeptide sequences

Protein-protein interaction (PPI) maps provide insight into cellular biology and have received considerable attention in the post-genomic era. While large-scale experimental approaches have generated large collections of experimentally determined PPIs, technical limitations preclude certain PPIs from detection. Recently, we demonstrated that yeast PPIs can be computationally predicted using re-occurring short polypeptide sequences between known interacting protein pairs. However, the computational requirements and low specificity made this method unsuitable for large-scale investigations. Here, we report an improved approach, which exhibits a specificity of approximately 99.95% and executes 16,000 times faster. Importantly, we report the first all-to-all sequence-based computational screen of PPIs in yeast, Saccharomyces cerevisiae in which we identify 29,589 high confidence interactions of approximately 2 x 10(7) possible pairs. Of these, 14,438 PPIs have not been previously reported and may represent novel interactions. In particular, these results reveal a richer set of membrane protein interactions, not readily amenable to experimental investigations. From the novel PPIs, a novel putative protein complex comprised largely of membrane proteins was revealed. In addition, two novel gene functions were predicted and experimentally confirmed to affect the efficiency of non-homologous end-joining, providing further support for the usefulness of the identified PPIs in biological investigations.     

Prediction of protein-protein interactions between Ralstonia solanacearum and Arabidopsis thaliana

Ralstonia solanacearum is a devastating bacterial pathogen that has an unusually wide host range. R. solanacearum, together with Arabidopsis thaliana, has become a model system for studying the molecular basis of plant-pathogen interactions. Protein-protein interactions (PPIs) play a critical role in the infection process, and some PPIs can initiate a plant defense response. However, experimental investigations have rarely addressed such PPIs. Using two computational methods, the interolog and the domain-based methods, we predicted 3,074 potential PPIs between 119 R. solanacearum and 1,442 A. thaliana proteins. Interestingly, we found that the potential pathogen-targeted proteins are more important in the A. thaliana PPI network. To facilitate further studies, all predicted PPI data were compiled into a database server called PPIRA (http://protein.cau.edu.cn/ppira/). We hope that our work will provide new insights for future research addressing the pathogenesis of R. solanacearum.     

Targeting protein-protein interactions for parasite control

Finding new drug targets for pathogenic infections would be of great utility for humanity, as there is a large need to develop new drugs to fight infections due to the developing resistance and side effects of current treatments. Current drug targets for pathogen infections involve only a single protein. However, proteins rarely act in isolation, and the majority of biological processes occur via interactions with other proteins, so protein-protein interactions (PPIs) offer a realm of unexplored potential drug targets and are thought to be the next-generation of drug targets. Parasitic worms were chosen for this study because they have deleterious effects on human health, livestock, and plants, costing society billions of dollars annually and many sequenced genomes are available. In this study, we present a computational approach that utilizes whole genomes of 6 parasitic and 1 free-living worm species and 2 hosts. The species were placed in orthologous groups, then binned in species-specific orthologous groups. Proteins that are essential and conserved among species that span a phyla are of greatest value, as they provide foundations for developing broad-control strategies. Two PPI databases were used to find PPIs within the species specific bins. PPIs with unique helminth proteins and helminth proteins with unique features relative to the host, such as indels, were prioritized as drug targets. The PPIs were scored based on RNAi phenotype and homology to the PDB (Protein DataBank). EST data for the various life stages, GO annotation, and druggability were also taken into consideration. Several PPIs emerged from this study as potential drug targets. A few interactions were supported by co-localization of expression in M. incognita (plant parasite) and B. malayi (H. sapiens parasite), which have extremely different modes of parasitism. As more genomes of pathogens are sequenced and PPI databases expanded, this methodology will become increasingly applicable.     

In silico discovery and experimental validation of new protein-protein interactions

We introduce a framework for predicting novel protein-protein interactions (PPIs), based on Fisher's method for combining probabilities of predictions that are based on different data sources, such as the biomedical literature, protein domain and mRNA expression information. Our method compares favorably to our previous method based on text-mining alone and other methods such as STRING. We evaluated our algorithms through the prediction of experimentally found protein interactions underlying Muscular Dystrophy, Huntington's Disease and Polycystic Kidney Disease, which had not yet been recorded in protein-protein interaction databases. We found a 1.74-fold increase in the mean average prediction precision for dysferlin and a 3.09-fold for huntingtin when compared to STRING. The top 10 of predicted interaction partners of huntingtin were analysed in depth. Five were identified previously, and the other five were new potential interaction partners. The full matrix of human protein pairs and their prediction scores are available for download. Our framework can be extended to predict other types of relationships such as proteins in a complex, pathway or related disease mechanisms.     

Protein complex finding and ranking: An application to Alzheimer’s disease

Protein complexes are known to play a major role in controlling cellular activity in a living being. Identifying complexes from raw protein–protein interactions (PPIs) is an important area of research. Earlier work has been limited mostly to yeast and a few other model organisms. Such protein complex identification methods, when applied to large human PPIs often give poor performance. We introduce a novel method called ComFiR to detect such protein complexes and further rank diseased complexes based on a query disease. We have shown that it has better performance in identifying protein complexes from human PPI data. This method is evaluated in terms of positive predictive value, sensitivity and accuracy. We have introduced a ranking approach and showed its application on Alzheimer’s disease.