Global Geometric Affinity for Revealing High Fidelity Protein Interaction Network

Abstract

Protein-protein interaction (PPI) network analysis presents an essential role in understanding the functional relationship among proteins in a living biological system. Despite the success of current approaches for understanding the PPI network, the large fraction of missing and spurious PPIs and a low coverage of complete PPI network are the sources of major concern. In this paper, based on the diffusion process, we propose a new concept of global geometric affinity and an accompanying computational scheme to filter the uncertain PPIs, namely, reduce the spurious PPIs and recover the missing PPIs in the network. The main concept defines a diffusion process in which all proteins simultaneously participate to define a similarity metric (global geometric affinity (GGA)) to robustly reflect the internal connectivity among proteins. The robustness of the GGA is attributed to propagating the local connectivity to a global representation of similarity among proteins in a diffusion process. The propagation process is extremely fast as only simple matrix products are required in this computation process and thus our method is geared toward applications in high-throughput PPI networks. Furthermore, we proposed two new approaches that determine the optimal geometric scale of the PPI network and the optimal threshold for assigning the PPI from the GGA matrix. Our approach is tested with three protein-protein interaction networks and performs well with significant random noises of deletions and insertions in true PPIs. Our approach has the potential to benefit biological experiments, to better characterize network data sets, and to drive new discoveries.

Protein Complexes Discovery Based on Protein-Protein Interaction Data via a Regularized Sparse Generative Network Model

“…different from previous approaches (e.g., MCode, SPICi and MINE) which detect complexes based on local neighborhood density, RSGNM is a global topological structure-based method. Since RSGNM takes the global structure of PPI network into account, it can make the best use of the pieces of local information simultaneously and hence can effectively handle the noisy interactions inherited in the high-throughput PPI networks [63]; 2. since our RSGNM is based on generative network model, it can handle the resolution limit problem in [42], [67], that is, RSGNM is able to discover complexes with any size. While, approach such as CFinder neglects complexes that really exist in the organisms but with size smaller than k, the k-clique size parameter in CFinder; 3. previous methods (e.g., MCL, RNSC, MCode, and SPICi) that do not support for detecting overlapping complexes cannot capture the real structure of complexes accurately in PPI networks in which some proteins usually belong to multiple complexes.…”

“…Here, we use (16) to initialize propensity matrix because of the assumption that similar proteins in the PPI network need to belong to the same complex, and ðD À1 W Þ 2 is a similarity matrix among proteins computed using global geometric affinity (GGA) with setting optimal propagation step to 2 [63]. Note that here the number of complexes parameter K 1 K þ n Àn is usually larger than the true number of complexes in corresponding PPI network.…”

Identification of DNA-Binding and Protein-Binding Proteins Using Enhanced Graph Wavelet Features

“…For example, Kumar et al developed DNAbinder [35] for predicting DNA-binding proteins from their amino acid sequence using various compositional features of proteins and exploited SVM classifier to classify DNA-binding proteins or not. Protein networks have been widely used to predict protein-protein interactions [36], [37], [38], [39], [40], [41], [42], [43], [44], which deal with a large number of interacting pairs in a single experiment and are more efficient than the former methods when the number of proteins is in thousands [38], [45], which are limited in preciseness of characterization yet [38].…”

“…For instance, the features cannot efficiently represent the physical properties of proteins, such as Van der Waals (VDW) force, and they do not consider the impact of the surrounding atoms. Motivated by the methods which also consider neighborhood information [39], [48], [49], the correlations between atomic features can be found. For example, Fig.…”

“…The enhanced graph wavelet features (EGWF) outperform the other three features, "Z" (Zhou and Yan features [46], [47]), "Z"þ"FA" (improved features by the way of additional new features), and "Z"þ"GW" (improved feature using graph wavelet alone), which indicates that the combination of two enhanced strategies most effectively improve Zhou et al's "Z" feature. Particularly, the comparison results between the weight matrix (WM) method that is used in Fang et al's [39] paper for predicting protein-protein interactions (PPI) via the topology of a PPI network, DNAbinder [35], and Zhou et al's method show that the graph wavelet is a useful tool to enhance the discrimination of the features. The main contributions of this work are presented as follows:…”