An Efficient Network Motif Discovery Approach for Co-Regulatory Networks

• Published in: IEEE access Vol. 6; pp. 14151 - 14158
• Main Authors: Luo, Jiawei, Ding, Lv, Liang, Cheng, Tu, Nguyen Hoang
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 01.01.2018; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Algorithms; Biological activity; Biomolecules; co-regulatory network motif; Computational complexity; Computer science; Enumeration; G-trie; Micro ribose nucleic acids (miRNAs); Molecular biophysics; Networks; Nodes; Nucleic acids; parallel processing; Regulators; Regulatory mechanisms (biology); Ribose; Sampling methods; Task analysis; transcription factor
• ISSN: 2169-3536; 2169-3536
• DOI: 10.1109/ACCESS.2018.2796565

Co-regulatory networks, which consist of transcription factors (TFs), micro ribose nucleic acids (miRNAs), and target genes, have provided new insight into biological processes, revealing complicated and comprehensive regulatory relationships between biomolecules. To uncover the key co-regulatory mechanisms between these biomolecules, the identification of co-regulatory motifs has become beneficial. However, due to high-computational complexity, it is a hard task to identify co-regulatory network motifs with more than four interacting nodes in large-scale co-regulatory networks. To overcome this limitation, we propose an efficient algorithm, named large co-regulatory network motif (LCNM), to detect large co-regulatory network motifs. This algorithm is able to store a set of co-regulatory network motifs within a <inline-formula> <tex-math notation="LaTeX">G </tex-math></inline-formula>-tries structure. Moreover, we propose two ways to generate candidate motifs. For three- or four-interacting-node motifs, LCNM is able to generate all different types of motif through an enumeration method. For larger network motifs, we adopt a sampling method to generate candidate co-regulatory motifs. The experimental results demonstrate that LCNM cannot only improve the computational performance in exhaustive identification of all of the three- or four-node motifs but can also identify co-regulatory network motifs with a maximum of eight nodes. In addition, we implement a parallel version of our LCNM algorithm to further accelerate the motif detection process.