Chern Han Yong

Methods for protein complex prediction and their contributions towards understanding the organisation, function and dynamics of complexes

Complexes of physically interacting proteins constitute fundamental functional units responsible for driving biological processes within cells. A faithful reconstruction of the entire set of complexes is therefore essential to understand the functional organisation of cells. In this review, we discuss the key contributions of computational methods developed till date (approximately between 2003 and 2015) for identifying complexes from the network of interacting proteins (PPI network). We evaluate in depth the performance of these methods on PPI datasets from yeast, and highlight their limitations and challenges, in particular at detecting sparse and small or sub‐complexes and discerning overlapping complexes. We describe methods for integrating diverse information including expression profiles and 3D structures of proteins with PPI networks to understand the dynamics of complex formation, for instance, of time‐based assembly of complex subunits and formation of fuzzy complexes from intrinsically disordered proteins. Finally, we discuss methods for identifying dysfunctional complexes in human diseases, an application that is proving invaluable to understand disease mechanisms and to discover novel therapeutic targets. We hope this review aptly commemorates a decade of research on computational prediction of complexes and constitutes a valuable reference for further advancements in this exciting area.

Stringent DDI-based Prediction of H. sapiens-M. tuberculosis H37Rv Protein-Protein Interactions

BackgroundH. sapiens-M. tuberculosis H37Rv protein-protein interaction (PPI) data are very important information to illuminate the infection mechanism of M. tuberculosis H37Rv. But current H. sapiens-M. tuberculosis H37Rv PPI data are very scarce. This seriously limits the study of the interaction between this important pathogen and its host H. sapiens. Computational prediction of H. sapiens-M. tuberculosis H37Rv PPIs is an important strategy to fill in the gap. Domain-domain interaction (DDI) based prediction is one of the frequently used computational approaches in predicting both intra-species and inter-species PPIs. However, the performance of DDI-based host-pathogen PPI prediction has been rather limited.ResultsWe develop a stringent DDI-based prediction approach with emphasis on (i) differences between the specific domain sequences on annotated regions of proteins under the same domain ID and (ii) calculation of the interaction strength of predicted PPIs based on the interacting residues in their interaction interfaces.We compare our stringent DDI-based approach to a conventional DDI-based approach for predicting PPIs based on gold standard intra-species PPIs and coherent informative Gene Ontology terms assessment. The assessment results show that our stringent DDI-based approach achieves much better performance in predicting PPIs than the conventional approach. Using our stringent DDI-based approach, we have predicted a small set of reliable H. sapiens-M. tuberculosis H37Rv PPIs which could be very useful for a variety of related studies.We also analyze the H. sapiens-M. tuberculosis H37Rv PPIs predicted by our stringent DDI-based approach using cellular compartment distribution analysis, functional category enrichment analysis and pathway enrichment analysis. The analyses support the validity of our prediction result. Also, based on an analysis of the H. sapiens-M. tuberculosis H37Rv PPI network predicted by our stringent DDI-based approach, we have discovered some important properties of domains involved in host-pathogen PPIs. We find that both host and pathogen proteins involved in host-pathogen PPIs tend to have more domains than proteins involved in intra-species PPIs, and these domains have more interaction partners than domains on proteins involved in intra-species PPI.ConclusionsThe stringent DDI-based prediction approach reported in this work provides a stringent strategy for predicting host-pathogen PPIs. It also performs better than a conventional DDI-based approach in predicting PPIs. We have predicted a small set of accurate H. sapiens-M. tuberculosis H37Rv PPIs which could be very useful for a variety of related studies.

Supervised maximum-likelihood weighting of composite protein networks for complex prediction

BackgroundProtein complexes participate in many important cellular functions, so finding the set of existent complexes is essential for understanding the organization and regulation of processes in the cell. With the availability of large amounts of high-throughput protein-protein interaction (PPI) data, many algorithms have been proposed to discover protein complexes from PPI networks. However, such approaches are hindered by the high rate of noise in high-throughput PPI data, including spurious and missing interactions. Furthermore, many transient interactions are detected between proteins that are not from the same complex, while not all proteins from the same complex may actually interact. As a result, predicted complexes often do not match true complexes well, and many true complexes go undetected.ResultsWe address these challenges by integrating PPI data with other heterogeneous data sources to construct a composite protein network, and using a supervised maximum-likelihood approach to weight each edge based on its posterior probability of belonging to a complex. We then use six different clustering algorithms, and an aggregative clustering strategy, to discover complexes in the weighted network. We test our method on Saccharomyces cerevisiae and Homo sapiens, and show that complex discovery is improved: compared to previously proposed supervised and unsupervised weighting approaches, our method recalls more known complexes, achieves higher precision at all recall levels, and generates novel complexes of greater functional similarity. Furthermore, our maximum-likelihood approach allows learned parameters to be used to visualize and evaluate the evidence of novel predictions, aiding human judgment of their credibility.ConclusionsOur approach integrates multiple data sources with supervised learning to create a weighted composite protein network, and uses six clustering algorithms with an aggregative clustering strategy to discover novel complexes. We show improved performance over previous approaches in terms of precision, recall, and number and quality of novel predictions. We present and visualize two novel predicted complexes in yeast and human, and find external evidence supporting these predictions.

Discovery of small protein complexes from PPI networks with size-specific supervised weighting

The prediction of small complexes (consisting of two or three distinct proteins) is an important and challenging subtask in protein complex prediction from protein-protein interaction (PPI) networks. The prediction of small complexes is especially susceptible to noise (missing or spurious interactions) in the PPI network, while smaller groups of proteins are likelier to take on topological characteristics of real complexes by chance.We propose a two-stage approach, SSS and Extract, for discovering small complexes. First, the PPI network is weighted by size-specific supervised weighting (SSS), which integrates heterogeneous data and their topological features with an overall topological isolatedness feature. SSS uses a naive-Bayes maximum-likelihood model to weight the edges with two posterior probabilities: that of being in a small complex, and of being in a large complex. The second stage, Extract, analyzes the SSS-weighted network to extract putative small complexes and scores them by cohesiveness-weighted density, which incorporates both small-co-complex and large-co-complex weights of edges within and surrounding the complexes.We test our approach on the prediction of yeast and human small complexes, and demonstrate that our approach attains higher precision and recall than some popular complex prediction algorithms. Furthermore, our approach generates a greater number of novel predictions with higher quality in terms of functional coherence.

Decomposing PPI networks for complex discovery

BackgroundProtein complexes are important for understanding principles of cellular organization and functions. With the availability of large amounts of high-throughput protein-protein interactions (PPI), many algorithms have been proposed to discover protein complexes from PPI networks. However, existing algorithms generally do not take into consideration the fact that not all the interactions in a PPI network take place at the same time. As a result, predicted complexes often contain many spuriously included proteins, precluding them from matching true complexes.ResultsWe propose two methods to tackle this problem: (1) The localization GO term decomposition method: We utilize cellular component Gene Ontology (GO) terms to decompose PPI networks into several smaller networks such that the proteins in each decomposed network are annotated with the same cellular component GO term. (2) The hub removal method: This method is based on the observation that hub proteins are more likely to fuse clusters that correspond to different complexes. To avoid this, we remove hub proteins from PPI networks, and then apply a complex discovery algorithm on the remaining PPI network. The removed hub proteins are added back to the generated clusters afterwards. We tested the two methods on the yeast PPI network downloaded from BioGRID. Our results show that these methods can improve the performance of several complex discovery algorithms significantly. Further improvement in performance is achieved when we apply them in tandem.ConclusionsThe performance of complex discovery algorithms is hindered by the fact that not all the interactions in a PPI network take place at the same time. We tackle this problem by using localization GO terms or hubs to decompose a PPI network before complex discovery, which achieves considerable improvement.

From the static interactome to dynamic protein complexes: Three challenges.

Protein interactions and complexes behave in a dynamic fashion, but this dynamism is not captured by interaction screening technologies, and not preserved in protein-protein interaction (PPI) networks. The analysis of static interaction data to derive dynamic protein complexes leads to several challenges, of which we identify three. First, many proteins participate in multiple complexes, leading to overlapping complexes embedded within highly-connected regions of the PPI network. This makes it difficult to accurately delimit the boundaries of such complexes. Second, many condition- and location-specific PPIs are not detected, leading to sparsely-connected complexes that cannot be picked out by clustering algorithms. Third, the majority of complexes are small complexes (made up of two or three proteins), which are extra sensitive to the effects of extraneous edges and missing co-complex edges. We show that many existing complex-discovery algorithms have trouble predicting such complexes, and show that our insight into the disparity between the static interactome and dynamic protein complexes can be used to improve the performance of complex discovery.

Prediction of problematic complexes from PPI networks: sparse, embedded, and small complexes

BackgroundThe prediction of protein complexes from high-throughput protein-protein interaction (PPI) data remains an important challenge in bioinformatics. Three groups of complexes have been identified as problematic to discover. First, many complexes are sparsely connected in the PPI network, and do not form dense clusters that can be derived by clustering algorithms. Second, many complexes are embedded within highly-connected regions of the PPI network, which makes it difficult to accurately delimit their boundaries. Third, many complexes are small (composed of two or three distinct proteins), so that traditional topological markers such as density are ineffective.ResultsWe have previously proposed three approaches to address these challenges. First, Supervised Weighting of Composite Networks (SWC) integrates diverse data sources with supervised weighting, and successfully fills in missing co-complex edges in sparse complexes to allow them to be predicted. Second, network decomposition (DECOMP) splits the PPI network into spatially- and temporally-coherent subnetworks, allowing complexes embedded within highly-connected regions to be more clearly demarcated. Finally, Size-Specific Supervised Weighting (SSS) integrates diverse data sources with supervised learning to weight edges in a size-specific manner—of being in a small complex versus a large complex—and improves the prediction of small complexes. Here we integrate these three approaches into a single system. We test the integrated approach on the prediction of yeast and human complexes, and show that it outperforms SWC, DECOMP, or SSS when run individually, achieving the highest precision and recall levels.ConclusionThree groups of protein complexes remain challenging to predict from PPI data: sparse complexes, embedded complexes, and small complexes. Our previous approaches have addressed each of these challenges individually, through data integration, PPI-network decomposition, and supervised learning. Here we integrate these approaches into a single complex-discovery system, which improves the prediction of all three types of challenging complexes. With our approach, protein complexes can be more accurately and comprehensively predicted, allowing a clearer elucidation of the modular machinery of the cell.ReviewersThis article was reviewed by Prof. Masanori Arita and Dr. Yang Liu (nominated by Prof. Charles DeLisi).