具時間序列微陣列資料建構基因調控網路之研究

The study of constructing gene regulary network from time-series microarrays data

陳宗堯(Zong-Yao Chen)

指導教授 ： 陳大正

國立虎尾科技大學/管理學院/資訊管理研究所/碩士(2009年)

摘要

本研究主要探討對具時間序列生物微陣列晶片的資料於基因調控網路之建構進行研究，亦即如何建立一個適當的基因調控網路(gene regulatory networks, GRNs)模型的方法，其能正確的由生物晶片內具時間序列的資料來解釋文獻上已經定義的幾種基因表現型態之間的關係，包含「正向調控關係」、「負向調控關係」、「正向時間延遲調控關係」之特徵，希冀可以提供準確率最高且方法更為簡易快速的方法來找出微陣列上的基因調控組合。由於微陣列晶片技術的蓬勃發展，伴隨而來的是龐大的基因表現時間序列資料。生物學家們嘗試著要從這龐大的數據中找出基因間彼此的關係，重建基因調控網路，此乃是公認的難題。因此，要如何準確且迅速的重建GRNs並充分表達其之間的關係，將會是生物晶片分析領域中一門非常值得研究的課題。在過去的此領域的文獻上，發現文獻上的方法皆無法準確的由具備時間序列的微陣列資料中探勘出上述三類不同的調控關係，此乃過去研究之不足處。
因此本研究提出一套基因調控網路的分析方法，利用核平滑(KS)合理的將雜訊的影響程度降低，再將平滑後的基因表現曲線進行離散快速傅利葉轉換(DFFT)萃取出基因表現序列之特徵，以提升兩兩基因間調控關係預測之準確率，並應用動態時間軸校準(DTW)來衡量兩兩基因間的調控機制是否具有延遲之現象，藉此達到快速且有效的重建基因網路，並能夠充分表達期間之關係。而此方法不僅可以針對過去文獻所提方法不足之處加以補強，更可表達更詳盡的調控關係資訊，以利於生物學家們進行調控關係之探討，進一步提供生物學家們一個快速且準確預測基因調控網路的捷徑。

關鍵字

基因調控網路(GRNs) ； 基因微陣列資料延遲性基因調控 ； 離散快速傅利葉轉換(DFFT) ； 動態時間軸校準(DTW)

並列摘要

This project is to investigate the issue of reconstructing gene regulatory networks (GRNs) from time-series microarray data. Four typical regulatory relationships depicted in the literature are active regulation without time delay, suppress regulation without time delay, active regulation with time delay, and suppress regulation with time delay. It is to develop the sequential approaches to mine out the regulatory relationship between any pairs of genes with highly accuracy rate. Due to the invention of the microarray technology, the tremendous of gene express time-series data are able to be obtained more easily. The biologists are trying to find out the exact regulatory relationship between to pairs of genes so as to reconstruct the gene regulatory networks. It is known GRNs is a difficult research work since it is hard to develop good approaches which can predict the regulatory relationships between any pairs of genes accurately. So, how to mine out the regulatory relationships between genes effectively and efficiently become the most important issues in the research area of biological microaray analysis. In the literature, it seems that no any approach can be provided to mine out the different four types of gene regulatory relationships. In this study, several tasks are to be achieved. It includes that the data noise can be processed adequately so that the gene expression profiles can be formed by the proposed approaches more accurately. Based on the gene expression profiles, the important features of the profile can be extracted. While the important features of profile are obtained, the gene regulatory relationship can be dig out by the proposed approach more perfectly. Moreover, The time-delay issue is included in the studies of gene regulation. The complete approaches are proposed and developed for mining three types of gene regulatory relationships. It is wished that the proposed approach in this study can improve the prediction accurate and also enhance the researchers’ interests of this area.

並列關鍵字

GRNs ； Time-series data ； microarray ； time delay

參考文獻

[1] Tobin, F.L., Damian-Lordache, V., and Greller, L.D. (1999), “Towards the reconstruction of gene regulatory networks,” Modeling and Simulation of Microsystems, 99, pp. 49-52.

[2] D'haeseleer, P., Wen, X., Fuhrman, S., and Somogyi, R. (1999), “Linear modeling of mRNA expression levels during CNS development and injury,” Proceeding of Pacific symposium on Biocomputing, pp. 41-52

[3] Xu, S., Lam, J., Ho, D., and Zou, Y. (2005), “Delay-dependent exponential stability for a class of neural networks with time delays,” Journal of Computational and Applied Mathematics, 183, pp.16-28.

[4] 張志方，「利用微陣列資料分析於基因調控網路之建構與預測」，國立中央大學資訊管理研究所碩士論文，2004。

[5] 張雅芳、黃正仲，「微陣列生物科技」，科學發展月刊，第38期：34-41，2004。