以社群多樣性影響力量度探採基因交互網絡之潛在調控成員

Social-diversity-based Influence Measurements for Mining Putative Regulators in Gene Interaction Networks

黃珮瑩(Pei-Ying Huang)

指導教授 ： 高成炎

國立臺灣大學/電機資訊學院/資訊工程學研究所/博士(2015年)

https://doi.org/10.6342/NTU.2015.11046

摘要

無資料

關鍵字

社群網路 ； 社群多樣性 ； 影響力傳輸 ； 基因交互作用網路 ； 社群多樣性影響力量度

並列摘要

We address the problem of identification of important nodes in the networks. For solving the problem, we propose four social-diversity-dependent schemes to identify important nodes via measuring the influence scores of nodes. They differ in the calculation of social diversities of the mediators. The prior model is based on the community structure. The zero-one spread and weighted spread are based on the static influence propagation while iter-weighted spread considers the dynamic influence spread.
Our findings on synthetic networks suggest that the social diversities of the mediators may play an important role in the identification of important nodes of various influence levels. Comparative analysis shows that iter-weighted spread is superior to our other three methods and PageRank, which implies that dynamic influence propagation may have an effect on discrimination of important nodes. It suggests that the pattern of the influence propagation should be updated dynamically to reflect the flow of influence spread to better capture the rapidly changing dynamics of networks.
Inspired by the observations on synthetic networks, we then apply our proposed method to two real-world networks: online social networks (e.g., Twitter) and protein-protein interaction (PPI) networks (e.g., yeast). On Twitter data, we employ iter-weighted spread to identify the influencers. Our results show that iter-weighted spread has a similar performance with PageRank for the high ranked users, while has better results than PageRank for middle ranked users.
On yeast data, we proposed a method named Networked Gene Ranker (NGR) integrating gene expression, social diversity and dynamic influence propagation to identify putative candidate genes in yeast PPI networks. Our results on the datasets of AmiGO meiotic genes reveal an interesting observation, node centrality measures perform better than other methods considering the prestige of the mediator. The results on DEG essential genes shows that, in general, NGR performs better than the existing methods.
Therefore, we conclude that both of the key mechanisms (i.e., social diversity and dynamic influence propagation) contribute to the detection and discrimination of influencers of difference influence levels in networks (e.g., social networks and PPI networks). Our proposed scheme is therefore practical and feasible to be deployed in the real world.

並列關鍵字

social network ； social diversity ； influence propagation ； protein-protein interaction network ； social-diversity-based influence measure

參考文獻

1. Freeman LC: Centrality in social networks: conceptual clarification. Social Networks 1978, 1(3):215-239.

2. Bonacich P: Power and Centrality: A Family of Measures. American Journal of Sociology 1987, 92(5):1170-1182.

3. Borgatti SP: Centrality and network flow. Social Networks 2005, 27(1):55-71.

4. Borgatti SP, Carley KM, Krackhardt D: On the robustness of centrality measures under conditions of imperfect data. Social Networks 2006, 28(2):124-136.

5. Pinski G, Narin F: Citation influence for journal aggregates of scientific publications: Theory, with application to the literature of physics. Information Processing & Management 1976, 12(5):297-312.