Predicting and validating protein interactions using network structure

Functional selectivity of Receptor Tyrosine Kinases regulates distinct cellular outputs

Functional selectivity refers to the activation of differential signalling and cellular outputs downstream of the same membrane-bound receptor when activated by two or more different ligands. Functional selectivity has been described and extensively studied for G-protein Coupled Receptors (GPCRs), leading to specific therapeutic options for dysregulated GPCRs functions. However, studies regarding the functional selectivity of Receptor Tyrosine Kinases (RTKs) remain sparse. Here, we will summarize recent data about RTK functional selectivity focusing on how the nature and the amount of RTK ligands and the crosstalk of RTKs with other membrane proteins regulate the specificity of RTK signalling. In addition, we will discuss how structural changes in RTKs upon ligand binding affects selective signalling pathways. Much remains to be known about the integration of different signals affecting RTK signalling specificity to orchestrate long-term cellular outcomes. Recent advancements in omics, specifically quantitative phosphoproteomics, and in systems biology methods to study, model and integrate different types of large-scale omics data have increased our ability to compare several signals affecting RTK functional selectivity in a global, system-wide fashion. We will discuss how such methods facilitate the exploration of important signalling hubs and enable data-driven predictions aiming at improving the efficacy of therapeutics for diseases like cancer, where redundant RTK signalling pathways often compromise treatment efficacy.

Deciphering the Host-Pathogen Interactome of the Wheat-Common Bunt System: A Step towards Enhanced Resilience in Next Generation Wheat

Common bunt, caused by two fungal species, Tilletia caries and Tilletia laevis, is one of the most potentially destructive diseases of wheat. Despite the availability of synthetic chemicals against the disease, organic agriculture relies greatly on resistant cultivars. Using two computational approaches—interolog and domain-based methods—a total of approximately 58 M and 56 M probable PPIs were predicted in T. aestivum–T. caries and T. aestivum–T. laevis interactomes, respectively. We also identified 648 and 575 effectors in the interactions from T. caries and T. laevis, respectively. The major host hubs belonged to the serine/threonine protein kinase, hsp70, and mitogen-activated protein kinase families, which are actively involved in plant immune signaling during stress conditions. The Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis of the host proteins revealed significant GO terms (O-methyltransferase activity, regulation of response to stimulus, and plastid envelope) and pathways (NF-kappa B signaling and the MAPK signaling pathway) related to plant defense against pathogens. Subcellular localization suggested that most of the pathogen proteins target the host in the plastid. Furthermore, a comparison between unique T. caries and T. laevis proteins was carried out. We also identified novel host candidates that are resistant to disease. Additionally, the host proteins that serve as transcription factors were also predicted.

Deciphering the Crosstalk Mechanisms of Wheat-Stem Rust Pathosystem: Genome-Scale Prediction Unravels Novel Host Targets

Triticum aestivum (wheat), a major staple food grain, is affected by various biotic stresses. Among these, fungal diseases cause about 15-20% of yield loss, worldwide. In this study, we performed a comparative analysis of protein-protein interactions between two Puccinia graminis races (Pgt 21-0 and Pgt Ug99) that cause stem (black) rust in wheat. The available molecular techniques to study the host-pathogen interaction mechanisms are expensive and labor-intensive. We implemented two computational approaches (interolog and domain-based) for the prediction of PPIs and performed various functional analysis to determine the significant differences between the two pathogen races. The analysis revealed that T. aestivum-Pgt 21-0 and T. aestivum-Pgt Ug99 interactomes consisted of ∼90M and ∼56M putative PPIs, respectively. In the predicted PPIs, we identified 115 Pgt 21-0 and 34 Pgt Ug99 potential effectors that were highly involved in pathogen virulence and development. Functional enrichment analysis of the host proteins revealed significant GO terms and KEGG pathways such as O-methyltransferase activity (GO:0008171), regulation of signal transduction (GO:0009966), lignin metabolic process (GO:0009808), plastid envelope (GO:0009526), plant-pathogen interaction pathway (ko04626), and MAPK pathway (ko04016) that are actively involved in plant defense and immune signaling against the biotic stresses. Subcellular localization analysis anticipated the host plastid as a primary target for pathogen attack. The highly connected host hubs in the protein interaction network belonged to protein kinase domain including Ser/Thr protein kinase, MAPK, and cyclin-dependent kinase. We also identified 5,577 transcription factors in the interactions, associated with plant defense during biotic stress conditions. Additionally, novel host targets that are resistant to stem rust disease were also identified. The present study elucidates the functional differences between Pgt 21-0 and Pgt Ug99, thus providing the researchers with strain-specific information for further experimental validation of the interactions, and the development of durable, disease-resistant crop lines.

Protein complex analysis: From raw protein lists to protein interaction networks

The elucidation of molecular interaction networks is one of the pivotal challenges in the study of biology. Affinity purification-mass spectrometry and other co-complex methods have become widely employed experimental techniques to identify protein complexes. These techniques typically suffer from a high number of false negatives and false positive contaminants due to technical shortcomings and purification biases. To support a diverse range of experimental designs and approaches, a large number of computational methods have been proposed to filter, infer and validate protein interaction networks from experimental pull-down MS data. Nevertheless, this expansion of available methods complicates the selection of the most optimal ones to support systems biology-driven knowledge extraction. In this review, we give an overview of the most commonly used computational methods to process and interpret co-complex results, and we discuss the issues and unsolved problems that still exist within the field. © 2015 Wiley Periodicals, Inc. Mass Spec Rev 36:600-614, 2017.

Detecting reliable non interacting proteins (NIPs) significantly enhancing the computational prediction of protein-protein interactions using machine learning methods

Protein-protein interactions (PPIs) play a vital role in most biological processes. Hence their comprehension can promote a better understanding of the mechanisms underlying living systems. However, besides the cost and the time limitation involved in the detection of experimentally validated PPIs, the noise in the data is still an important issue to overcome. In the last decade several in silico PPI prediction methods using both structural and genomic information were developed for this purpose. Here we introduce a unique validation approach aimed to collect reliable non interacting proteins (NIPs). Thereafter the most relevant protein/protein-pair related features were selected. Finally, the prepared dataset was used for PPI classification, leveraging the prediction capabilities of well-established machine learning methods. Our best classification procedure displayed specificity and sensitivity values of 96.33% and 98.02%, respectively, surpassing the prediction capabilities of other methods, including those trained on gold standard datasets. We showed that the PPI/NIP predictive performances can be considerably improved by focusing on data preparation.

LocFuse: human protein-protein interaction prediction via classifier fusion using protein localization information

Protein-protein interaction (PPI) detection is one of the central goals of functional genomics and systems biology. Knowledge about the nature of PPIs can help fill the widening gap between sequence information and functional annotations. Although experimental methods have produced valuable PPI data, they also suffer from significant limitations. Computational PPI prediction methods have attracted tremendous attentions. Despite considerable efforts, PPI prediction is still in its infancy in complex multicellular organisms such as humans. Here, we propose a novel ensemble learning method, LocFuse, which is useful in human PPI prediction. This method uses eight different genomic and proteomic features along with four types of different classifiers. The prediction performance of this classifier selection method was found to be considerably better than methods employed hitherto. This confirms the complex nature of the PPI prediction problem and also the necessity of using biological information for classifier fusion. The LocFuse is available at: http://lbb.ut.ac.ir/Download/LBBsoft/LocFuse.

BIOLOGICAL SIGNIFICANCE

The results revealed that if we divide proteome space according to the cellular localization of proteins, then the utility of some classifiers in PPI prediction can be improved. Therefore, to predict the interaction for any given protein pair, we can select the most accurate classifier with regard to the cellular localization information. Based on the results, we can say that the importance of different features for PPI prediction varies between differently localized proteins; however in general, our novel features, which were extracted from position-specific scoring matrices (PSSMs), are the most important ones and the Random Forest (RF) classifier performs best in most cases. LocFuse was developed with a user-friendly graphic interface and it is freely available for Linux, Mac OSX and MS Windows operating systems.

Global protein-protein interaction network of rice sheath blight pathogen

Rhizoctonia solani is the major pathogenic fungi of rice sheath blight. It is responsible for the most serious disease of rice (Oryza sativa L.) and causes significant yield losses in rice-growing countries. Identifying the protein-protein interaction (PPI) maps of R. solani can provide insights into the potential pathogenic mechanisms and assign putative functions to unknown genes. Here, we exploited a PPI map of R. solani anastomosis group 1 IA (AG-1 IA) based on the interolog and domain-domain interaction methods. We constructed a core subset of high-confidence protein networks consisting of 6705 interactions among 1773 proteins. The high quality of the network was revealed by comprehensive methods, including yeast two-hybrid experiments. Pathogenic interaction subnetwork, secreted proteins subnetwork, and mitogen-activated protein kinase (MAPK) cascade subnetwork and their interacting partners were constructed and analyzed. Moreover, to exactly predict the pathogenic factors, the expression levels of the interaction proteins were investigated by analyzing RNA sequences that consisted of samples from the entire infection progress. The PPIs offer an exceptionally rich source of data that can be used to understand the gene functions and biological processes of this serious disease at the system level.

in silico identification of protein-protein interactions in Silkworm, Bombyx mori

The Domesticated silkworm, Bombyx mori, an economically important insect has been used as a lepidopteran molecular model next only to Drosophila. Compared to the genomic information in silkworm, the protein-protein interaction data are limited. Therefore experimentally identified PPI maps from five model organisms such as E.coli, C.elegans, D.melanogaster, H. sapiens, S. cerevisiae were used to infer the PPI network of silkworm using the well-recognized Interlog based method. Among the 14623 silkworm proteins, 7736 protein-protein interaction pairs were predicted which include 2700 unique proteins of the silkworms. Using the iPfam interaction domains and the gene expression data, these predictions were validated. In that 625 PPI pairs of predicted network were associated with the iPfam domain-domain interactions and the random network has average of 9. In the gene expression method, the average PCC value of the predicted network and random network was 0.29 and 0.23100±0.00042 respectively. It reveals that the predicted PPI networks of silkworm are highly significant and reliable. This is the first PPI network for the silkworm which will provide a framework for deciphering the cellular processes governing key metabolic pathways in the silkworm, Bombyx mori and available at SilkPPI (http://210.212.197.30/SilkPPI/).

Computational Prediction of Protein-Protein Interaction Networks: Algo-rithms and Resources

Protein interactions play an important role in the discovery of protein functions and pathways in biological processes. This is especially true in case of the diseases caused by the loss of specific protein-protein interactions in the organism. The accuracy of experimental results in finding protein-protein interactions, however, is rather dubious and high throughput experimental results have shown both high false positive beside false negative information for protein interaction. Computational methods have attracted tremendous attention among biologists because of the ability to predict protein-protein interactions and validate the obtained experimental results. In this study, we have reviewed several computational methods for protein-protein interaction prediction as well as describing major databases, which store both predicted and detected protein-protein interactions, and the tools used for analyzing protein interaction networks and improving protein-protein interaction reliability.

Antibody i-Patch prediction of the antibody binding site improves rigid local antibody-antigen docking

Antibodies are a class of proteins indispensable for the vertebrate immune system. The general architecture of all antibodies is very similar, but they contain a hypervariable region which allows millions of antibody variants to exist, each of which can bind to different molecules. This binding malleability means that antibodies are an increasingly important category of biopharmaceuticals and biomarkers. We present Antibody i-Patch, a method that annotates the most likely antibody residues to be in contact with the antigen. We show that our predictions correlate with energetic importance and thus we argue that they may be useful in guiding mutations in the artificial affinity maturation process. Using our predictions as constraints for a rigid-body docking algorithm, we are able to obtain high-quality results in minutes. Our annotation method and re-scoring system for docking achieve their predictive power by using antibody-specific statistics. Antibody i-Patch is available from http://www.stats.ox.ac.uk/research/proteins/resources.

Complementing the Eukaryotic Protein Interactome

Protein interaction networks are important for the understanding of regulatory mechanisms, for the explanation of experimental data and for the prediction of protein functions. Unfortunately, most interaction data is available only for model organisms. As a possible remedy, the transfer of interactions to organisms of interest is common practice, but it is not clear when interactions can be transferred from one organism to another and, thus, the confidence in the derived interactions is low. Here, we propose to use a rich set of features to train Random Forests in order to score transferred interactions. We evaluated the transfer from a range of eukaryotic organisms to S. cerevisiae using orthologs. Directly transferred interactions to S. cerevisiae are on average only 24% consistent with the current S. cerevisiae interaction network. By using commonly applied filter approaches the transfer precision can be improved, but at the cost of a large decrease in the number of transferred interactions. Our Random Forest approach uses various features derived from both the target and the source network as well as the ortholog annotations to assign confidence values to transferred interactions. Thereby, we could increase the average transfer consistency to 85%, while still transferring almost 70% of all correctly transferable interactions. We tested our approach for the transfer of interactions to other species and showed that our approach outperforms competing methods for the transfer of interactions to species where no experimental knowledge is available. Finally, we applied our predictor to score transferred interactions to 83 targets species and we were able to extend the available interactome of B. taurus, M. musculus and G. gallus with over 40,000 interactions each. Our transferred interaction networks are publicly available via our web interface, which allows to inspect and download transferred interaction sets of different sizes, for various species, and at specified expected precision levels.

AVAILABILITY

http://services.bio.ifi.lmu.de/coin-db/.

Exploiting conformational ensembles in modeling protein-protein interactions on the proteome scale

Cellular functions are performed through protein-protein interactions; therefore, identification of these interactions is crucial for understanding biological processes. Recent studies suggest that knowledge-based approaches are more useful than "blind" docking for modeling at large scales. However, a caveat of knowledge-based approaches is that they treat molecules as rigid structures. The Protein Data Bank (PDB) offers a wealth of conformations. Here, we exploited an ensemble of the conformations in predictions by a knowledge-based method, PRISM. We tested "difficult" cases in a docking-benchmark data set, where the unbound and bound protein forms are structurally different. Considering alternative conformations for each protein, the percentage of successfully predicted interactions increased from ~26 to 66%, and 57% of the interactions were successfully predicted in an "unbiased" scenario, in which data related to the bound forms were not utilized. If the appropriate conformation, or relevant template interface, is unavailable in the PDB, PRISM could not predict the interaction successfully. The pace of the growth of the PDB promises a rapid increase of ensemble conformations emphasizing the merit of such knowledge-based ensemble strategies for higher success rates in protein-protein interaction predictions on an interactome scale. We constructed the structural network of ERK interacting proteins as a case study.

Incorporating Correlations among Gene Ontology Terms into Predicting Protein Functions

The authors describe a new strategy that has better prediction performance than previous methods, which gives additional insights about the importance of the dependence between functional terms when inferring protein function.

The importance of age and high degree, in protein-protein interaction networks

Here we present an in-depth analysis of the protein age patterns found in the edge and triangle subgraphs of the yeast protein-protein interaction network (PIN). We assess their statistical significance both according to what would be expected by chance given the node frequencies found in the yeast PIN, and also, for the case of triangles, given the age frequencies observed in the currently available pairwise data. We find that pairwise interactions between Old proteins are over-represented even when controlling for high degree, and triangle interactions between Old proteins are over-represented even when controlling for pairwise interaction frequencies. There is evidence for negative selection of interactions between Middle-aged and Old proteins within triangles, despite pairwise Middle-Old interactions being common. Most triangles consist solely of vertices with high degree. Our findings point towards an architecture of the yeast PIN that is highly heterogeneous, having connected clumps which contain a large number of interacting Old proteins along with selective age-dependent interaction patterns. Supplementary Material is available online (www.liebertonline.com/cmb).

Prediction and characterization of protein-protein interaction networks in swine

Background

Studying the large-scale protein-protein interaction (PPI) network is important in understanding biological processes. The current research presents the first PPI map of swine, which aims to give new insights into understanding their biological processes.

Results

We used three methods, Interolog-based prediction of porcine PPI network, domain-motif interactions from structural topology-based prediction of porcine PPI network and motif-motif interactions from structural topology-based prediction of porcine PPI network, to predict porcine protein interactions among 25,767 porcine proteins. We predicted 20,213, 331,484, and 218,705 porcine PPIs respectively, merged the three results into 567,441 PPIs, constructed four PPI networks, and analyzed the topological properties of the porcine PPI networks. Our predictions were validated with Pfam domain annotations and GO annotations. Averages of 70, 10,495, and 863 interactions were related to the Pfam domain-interacting pairs in iPfam database. For comparison, randomized networks were generated, and averages of only 4.24, 66.79, and 44.26 interactions were associated with Pfam domain-interacting pairs in iPfam database. In GO annotations, we found 52.68%, 75.54%, 27.20% of the predicted PPIs sharing GO terms respectively. However, the number of PPI pairs sharing GO terms in the 10,000 randomized networks reached 52.68%, 75.54%, 27.20% is 0. Finally, we determined the accuracy and precision of the methods. The methods yielded accuracies of 0.92, 0.53, and 0.50 at precisions of about 0.93, 0.74, and 0.75, respectively.

Conclusion

The results reveal that the predicted PPI networks are considerably reliable. The present research is an important pioneering work on protein function research. The porcine PPI data set, the confidence score of each interaction and a list of related data are available at (http://pppid.biositemap.com/).

Experimental and computational approaches for the study of calmodulin interactions

Ca(2+), a universal messenger in eukaryotes, plays a major role in signaling pathways that control many growth and developmental processes in plants as well as their responses to various biotic and abiotic stresses. Cellular changes in Ca(2+) in response to diverse signals are recognized by protein sensors that either have their activity modulated or that interact with other proteins and modulate their activity. Calmodulins (CaMs) and CaM-like proteins (CMLs) are Ca(2+) sensors that have no enzymatic activity of their own but upon binding Ca(2+) interact and modulate the activity of other proteins involved in a large number of plant processes. Protein-protein interactions play a key role in Ca(2+)/CaM-mediated in signaling pathways. In this review, using CaM as an example, we discuss various experimental approaches and computational tools to identify protein-protein interactions. During the last two decades hundreds of CaM-binding proteins in plants have been identified using a variety of approaches ranging from simple screening of expression libraries with labeled CaM to high-throughput screens using protein chips. However, the high-throughput methods have not been applied to the entire proteome of any plant system. Nevertheless, the data provided by these screens allows the development of computational tools to predict CaM-interacting proteins. Using all known binding sites of CaM, we developed a computational method that predicted over 700 high confidence CaM interactors in the Arabidopsis proteome. Most (>600) of these are not known to bind calmodulin, suggesting that there are likely many more CaM targets than previously known. Functional analyses of some of the experimentally identified Ca(2+) sensor target proteins have uncovered their precise role in Ca(2+)-mediated processes. Further studies on identifying novel targets of CaM and CMLs and generating their interaction network - "calcium sensor interactome" - will help us in understanding how Ca(2+) regulates a myriad of cellular and physiological processes.

A predicted protein-protein interaction network of the filamentous fungus Neurospora crassa

The filamentous fungus Neurospora crassa is a leading model organism for circadian clock studies. Computational identification of a protein-protein interaction (PPI) network (also known as an interactome) in N. crassa can provide new insights into the cellular functions of proteins. Using two well-established bioinformatics methods (the interolog method and the domain interaction-based method), we predicted 27,588 PPIs among 3006 N. crassa proteins. To the best of our knowledge, this is the first identified interactome for N. crassa, although it remains problematic because of incomplete interactions and false positives. In particular, the established PPI network has provided clues to further decipher the molecular mechanism of circadian rhythmicity. For instance, we found that clock-controlled genes (ccgs) are more likely to act as bottlenecks in the established PPI network. We also identified an important module related to circadian oscillators, and some functional unknown proteins in this module may serve as potential candidates for new oscillators. Finally, all predicted PPIs were compiled into a user-friendly database server (NCPI), which is freely available at .

Integrating diverse biological and computational sources for reliable protein-protein interactions

Background

Protein-protein interactions (PPIs) play important roles in various cellular processes. However, the low quality of current PPI data detected from high-throughput screening techniques has diminished the potential usefulness of the data. We need to develop a method to address the high data noise and incompleteness of PPI data, namely, to filter out inaccurate protein interactions (false positives) and predict putative protein interactions (false negatives).

Results

In this paper, we proposed a novel two-step method to integrate diverse biological and computational sources of supporting evidence for reliable PPIs. The first step, interaction binning or InterBIN, groups PPIs together to more accurately estimate the likelihood (Bin-Confidence score) that the protein pairs interact for each biological or computational evidence source. The second step, interaction classification or InterCLASS, integrates the collected Bin-Confidence scores to build classifiers and identify reliable interactions.

Conclusions

We performed comprehensive experiments on two benchmark yeast PPI datasets. The experimental results showed that our proposed method can effectively eliminate false positives in detected PPIs and identify false negatives by predicting novel yet reliable PPIs. Our proposed method also performed significantly better than merely using each of individual evidence sources, illustrating the importance of integrating various biological and computational sources of data and evidence.

Triangle network motifs predict complexes by complementing high-error interactomes with structural information

Background

A lot of high-throughput studies produce protein-protein interaction networks (PPINs) with many errors and missing information. Even for genome-wide approaches, there is often a low overlap between PPINs produced by different studies. Second-level neighbors separated by two protein-protein interactions (PPIs) were previously used for predicting protein function and finding complexes in high-error PPINs. We retrieve second level neighbors in PPINs, and complement these with structural domain-domain interactions (SDDIs) representing binding evidence on proteins, forming PPI-SDDI-PPI triangles.

Results

We find low overlap between PPINs, SDDIs and known complexes, all well below 10%. We evaluate the overlap of PPI-SDDI-PPI triangles with known complexes from Munich Information center for Protein Sequences (MIPS). PPI-SDDI-PPI triangles have ~20 times higher overlap with MIPS complexes than using second-level neighbors in PPINs without SDDIs. The biological interpretation for triangles is that a SDDI causes two proteins to be observed with common interaction partners in high-throughput experiments. The relatively few SDDIs overlapping with PPINs are part of highly connected SDDI components, and are more likely to be detected in experimental studies. We demonstrate the utility of PPI-SDDI-PPI triangles by reconstructing myosin-actin processes in the nucleus, cytoplasm, and cytoskeleton, which were not obvious in the original PPIN. Using other complementary datatypes in place of SDDIs to form triangles, such as PubMed co-occurrences or threading information, results in a similar ability to find protein complexes.

Conclusion

Given high-error PPINs with missing information, triangles of mixed datatypes are a promising direction for finding protein complexes. Integrating PPINs with SDDIs improves finding complexes. Structural SDDIs partially explain the high functional similarity of second-level neighbors in PPINs. We estimate that relatively little structural information would be sufficient for finding complexes involving most of the proteins and interactions in a typical PPIN.

Clustering algorithms for detecting functional modules in protein interaction networks

Protein-Protein Interaction (PPI) networks are believed to be important sources of information related to biological processes and complex metabolic functions of the cell. When studying the workings of a biological cell, it is useful to be able to detect known and predict still undiscovered protein complexes within the cell's PPI networks. Such predictions may be used as an inexpensive tool to direct biological experiments. The increasing amount of available PPI data necessitate a fast, accurate approach to biological complex identification. Because of its importance in the studies of protein interaction network, there are different models and algorithms in identifying functional modules in PPI networks. In this paper, we review some representative algorithms, focusing on the algorithms underlying the approaches and how the algorithms relate to each other. In particular, a comparison is given based on the property of the algorithms. Since the PPI network is noisy and still incomplete, some methods which consider other additional properties for preprocessing and purifying of PPI data are presented. We also give a discussion about the functional annotation and validation of protein complexes. Finally, new progress and future research directions are discussed from the computational viewpoint.