Prioritizing network communities

Abstract

Uncovering modular structure in networks is fundamental for advancing the understanding of complex systems in biology, physics, engineering, and technology. Community detection provides a way to computationally identify candidate modules as hypotheses, which then need to be experimentally validated. However, validation of detected communities requires expensive and time consuming experimental methods, such as mutagenesis in a wet biological laboratory. As a consequence only a limited number of communities can be experimentally validated, and it is thus important to determine which communities to select for downstream validation and experimentation. Here we develop CRANK, an automatic method for prioritizing network communities and identifying the most promising ones for further experimentation. CRANK efficiently evaluates robustness and magnitude of structural features of each community and then combines these features to obtain the community prioritization. CRANK can be used with any community detection method and scales to large networks. It needs only information provided by the network structure and does not require any additional metadata or labels. However, when available, CRANK can incorporate domain-specific information to further boost performance. Experiments on many diverseand important biological networks demonstrate that the proposed approach effectively prioritizes communities, yielding a near 50-fold improvement in community prioritization over a baseline ordering of detected communities. Taken together, CRANK represents a networkbased approach to identify high-quality communities, even for domains at the frontier of science where supervised meta information is not available.

DTI-Voodoo: machine learning over interaction networks and ontology-based background knowledge predicts drug–target interactions

“…As the difference in performance with different splitting schemes is quite large, we further evaluated additional drug–target interaction and drug–target affinity prediction methods that were trained and evaluated on other datasets. Following the results of MolTrans (Huang et al ., 2020), we reevaluated DeepDTI (Wen et al ., 2017), DeepDTA (Öztürk et al ., 2018), DeepConv-DTI (Lee et al ., 2019), and MolTrans itself on the BioSnap dataset (Zitnik et al ., 2018) and compared it to our “naïve” predictor as well as DTI-Voodoo (see Table 3). MolTrans was evaluated over the drug– target pair and the protein split; we were able to reproduce the MolTrans results (Table 3), showing a substantial difference based on the splitting scheme.…”

“…For comparative evaluation, we use the gold standard dataset introduced by Yamanishi et al . (2008) consisting of 1,923 interactions between 708 drugs and 1,512 proteins, and the BioSnap dataset (Zitnik et al ., 2018) which consists of 5,017 drug nodes, 2,324 gene nodes and 15,138 edges.…”

Predicting Biomedical Interactions With Higher-Order Graph Convolutional Networks

“…We conduct interaction prediction experiments on four publicly-available biomedical network datasets: BioSNAP-DTI [ 28 ]: DTI network contains 15,139 drug-target interactions between 5,018 drugs and 2,325 proteins. BioSNAP-DDI [ 28 ]: DDI network contains 48,514 drug-drug interactions between 1,514 drugs extracted from drug labels and biomedical literature. HuRI-PPI [ 2 ]: HI-III human PPI network contains 5,604 proteins and 23,322 interactions generated by multiple orthogonal high-throughput yeast two-hybrid screens.…”

“…BioSNAP-DTI [ 28 ]: DTI network contains 15,139 drug-target interactions between 5,018 drugs and 2,325 proteins.…”

Mobile Networks and Internet of Things Infrastructures to Characterize Smart Human Mobility

“…It is a location-based social networking website where users share their locations by checkingin. The friendship network is undirected and was collected using their public API [40], and consists of 196,591 nodes and 950,327 edges. We have collected a total of 6,442,890 check-ins of these users over the period of February 2009-October 2010.…”