NemoMap: Improved Motif-centric Network Motif Discovery Algorithm

Abstract

Network motif analysis has several applications in many different fields such as biological study and social network modeling, yet motif detection tools are still limited by the intensive computation. Currently, there are two categories for network motif detection method: network-centric and motif-centric approach. While most network-centric algorithms excel in enumerating all potential motifs of a given size, the runtime is infeasible for larger size of motifs. Researchers who are interested in larger motifs and have established a set of potential motif patterns could utilize motif-centric tools to check whether such patterns are truly network motifs by mapping them to the target network and counting their frequency. In the paper, we present NemoMap (Network Motif Mapping algorithm) which is an improvement of the motif-centric algorithm, GK (by Grochow and Kellis) and MODA (Motif Detection Algorithm). Experimental results on three different protein-protein interaction networks show that NemoMap is more efficient in mapping complex motif patterns, while GK and MODA is much faster in analyzing simpler patterns with fewer edges. We also compare the performance of NemoMap and ParaMODA (introduced previously to improve MODA), and the result shows that NemoMap yields better runtime due to the implementation of Grochow-Kellis' symmetry-breaking technique and the better node selection process.

NemoSuite: Web-based Network Motif Analytic Suite

“…Nemo program employs a network-centric approach using a NemoLib [15,16] that improves ESU (Enumerate SUbgraphs) algorithm [33]. NemoMapPy program follows a motif-centric method by implementing NemoMap [17] algorithm in Python and Flask framework. Figure 5 shows the NemoSuite homepage with Nemo and NemoMapPy program pages.…”

“…Focusing on detecting larger size of motifs, Grochow and Kellis introduced a motif-centric approach (GK method), which finds the frequency of each pattern in a given query set by mapping each query graph to all possible location in the input graph [24]. GK method, MODA [35], FASCIA [36], ISMAGS [31], ParaMODA [37], and NemoMap [17] are implementing this motif-centric approach. This approach was able to find motifs of up to 15 nodes with symmetrybreaking technique to significantly reduce isomorphic testing.…”

“…NemoMapPy, as an online tool, refined and upgraded the features of NemoMap [17] that improved the performance of ParaMODA [37], which also increased the efficiency of the motifcentric algorithm introduced by Grochow and Kellis (GK) [24], and MODA program [35]. In fact, NemoMap improved the performance of ParaMODA, MODA and GK method by modifying the symmetry-breaking method.…”

“…We tested the efficiency of NemoMapPy by comparing with the NemoMap, as NemoMap is known the fastest program with motifcentric approach [17] in most of cases, when not considering parallelization. We tested NemoMapPy with a graph of 1,018 nodes and 1,331 edges, the query graphs that has 9, 10, 11 nodes with average complexity.…”

“…Nemo program implements a network-centric approach using NemoLib [15,16]. NemoMapPy implements NemoMap [17] algorithm which is one of the motif-centric methods.…”