基於圖形群聚之基因網路重建演算法

Graph-Based Clustering Approaches for Gene Network Reconstruction

賴至祥(Jhih-Siang Lai)

指導教授 ： 陳中明

國立臺灣大學/工學院/醫學工程學研究所/碩士(2009年)

https://doi.org/10.6342/NTU.2009.01576

摘要

為了解生物基因間的調控關係，生物學家常利用干擾性核醣核酸(RNAi)，或是基因剔除(gene knockout)的方式來觀察生物系統的反應。資訊學家則嘗試利用演算法以mRNA隨時間變化的表現量曲線重建出可能的基因間調控關係。然而，基因間的調控包含許多階段，包括轉錄 (Transcription)、轉錄後修飾(Post-transcriptional modification) 、轉譯 (Translation) 、mRNA的降解 (mRNA degradation) 、轉譯後修飾 (Post-translational modification)等。這些階段都需要時間來反應，因此許多研究根據時間延遲的特徵，分析基因間的調控關係。
在這份研究中，我們使用兩個方法來重建基因網路。一個是Normalized Cuts
，以圖學方式試著將有功能性的基因調控網路分割出來。另一個方法則是PARE (Pattern Recognition Approach)演算法，一個以時間延遲(time-lagged)以及非線性特徵作為基因間調控關係的推論演算法。我們使用酵母菌的mRNA隨時間變化的表現量作為重建基因調控網路的分析材料，再以KEGG pathway資料庫、BIOGRID 交互影響資料庫與MIPS資料庫做為比較分析的參考。而從分析出的F score結果來看，我們的方法優於Kim等人所發展出的動態貝式網路。
最後，我們將方法應用到一個實際的例子，yox1與yhp1兩個基因皆剔除的酵母菌的生物晶片上，分析其mRNA隨時間變化的表現量。由於細胞每段時期間轉換機制尚未完全被了解，目前已知yox1與yhp1是以負回饋的機制控制細胞在G1時期的時間。我們成功地找到與細胞生命週期相關的調控網路，其中一個調控網路與細胞分裂相關。藉由這份應用結果，我們期望能夠探究出更多關於細胞生命週期中每個時期轉換間的調控機制。

關鍵字

基因網路 ； 正規切割 ； 時間延遲

並列摘要

To understand regulatory relationships between genes in real life. Biologists often use RNA interference (RNAi) or knockout genes to observe the response in the real life system. Informationists try to reconstruct regulatory relationship between genes from mRNA expression profile by algorithms or mathematic models. There are several phases involved in gene regulation such as transcription, post-transcriptional modifications, translation,
mRNA degradation and post-translational modifications .Time is essential for all these phases to be completed and many researches analyze regulation via these features.
In this study, we use two methods to reconstruct regulatory relationships between genes. One is a graph partition algorithm named Normalized Cuts for partitioning off genes into functional gene network. The other method, PARE (Pattern Recognition Approach), an algorithm based on time-lagged non-linear feature of the profile, is to infer regulation between genes. In addition, we use yeast microarray to construct gene regulatory networks and check results from KEGG pathway database, BIOGRID interaction database and MIPS database. Comparing our F score result with Dynamic Bayesian Network developed by Kim, et al., it shows that our method performs better than theirs.
Finally, we apply our method to a real case in yeast microarray in which yox1 and yhp1 are both deleted and we analyze its mRNA expression time profile. Although mechanisms between phases in cell cycle are not clear, yox1 and yhp1 are two genes known controlling duration of a cell in G1 phase by negative feedback. We successfully find networks associated with cell cycle and one of the networks is associated with cell mitosis. In the future, we hope to decipher more mechanisms between phases in cell cycle.

並列關鍵字

Gene Network ； Normalized Cuts ； Time Lag

參考文獻

[01] Chen T, He HL, Church GM. (1999) Modeling gene expression with differential equations. Pac Symp Biocomput. 4. 29-40.

[02] Ideker TE, Thorsson V, Karp RM. (2000) Discovery of regulatory interactions through perturbation: inference and experimental design. Pac Symp Biocomput. 5. 302-313.

[03] Friedman N, Linial M, Nachman I, Pe'er D. (2000) Using Bayesian networks to analyze expression data. J Comput Biol. 7. 601-620.

[04] Chen T, Filkov V, Skiena SS. (2001) Identifying gene regulatory networks from experimental data. Parallel Computing. 27. 141-162.

[05] Arkin A,Shen P, Ross J. (1997) A Test Case of Correlation Metric Construction of a Reaction Pathway from Measurements. Science . 277. 1275-1279.