Treating proteins as nodes and interactions between proteins as edges, we obtain a sparse protein-protein interaction network.
Minimal Screen 12 [ 46], a sparse matrix protein crystallization screen [ 47], was used to search for suitable conditions for crystallization of AVR2-b with the vapor diffusion hanging drop method at 22°C.
The sparse connectivity of protein-protein interaction data sets makes identification of functional modules challenging.
Considering the difficulty of dealing with sparse and noisy protein networks (You et al., 2010), our results represent a promising achievement and encouragement to further the investigation of network embedding techniques for topological prediction of candidate protein interactions.
Methods that remove such transitive correlations have been developed in recent years and include, for example, Direct Coupling Analysis (DCA or DI for Direct Information) (Weigt et al., 2009; Morcos et al., 2011), Protein Sparse Inverse COVariance (PSICOV) (Jones et al., 2012) and Gremlin's pseudolikelihood method (Kamisetty et al., 2013).
Such methods include direct coupling analysis (DCA or DI for direct information) (Morcos et al., 2011; Weigt et al., 2009), Protein Sparse Inverse COVariance (PSICOV) (Jones et al., 2012), a Bayesian network algorithm for disentangling direct from indirect dependencies between residues (Burger and van Nimwegen, 2010), the pseudolikelihood maximization DCA (plmDCA) method due to Ekeberg et al.
Annotations that appears in >40 proteins are sparse and include 'Zinc ion binding' (46 proteins), 'G-protein coupled receptor protein signaling' (42 proteins), and 'Calcium binding' (45 proteins).
Another limitation in our analysis is the relatively sparse coverage of current protein-protein interaction databases.
Our method works particularly well for distantly related proteins or proteins with sparse sequence profiles because of the effective integration of context-specific, structure and global information.
