以結構方程模型重建基因網路

Abstract

隨著生物基因研究持續進行，越來越多的基因被發現與疾病有直接關係，探索基因的功能已成為生物技術研究的重點項目。基因的運作並非單獨進行，而是透過一連串複雜的交互作用機制對生物體產生影響，且由於生物體的複雜性，大部分疾病與基因的關係不容易釐清。重建基因網路的目的就是為了分析基因之間互相調控的運作機制，進一步了解基因對生物體產生影響的運作細節。 受限於微陣列晶片(microarray)成本過高，生物實驗通常無法提供大量的連續觀測資料以重建基因網路。為發展一套利用少量觀測值有效重建基因網路的演算法，本論文選擇以結構方程模型為基本架構，提出了以線性動態因素模型來解釋基因間互動的演算法。除了可以測量的變數以外，我們在線性動態因素模型也納入了潛在因素(latent factors)以解釋那些對整個基因網路有所影響但我們沒納入的基因或分子。我們模擬了6個基因的基因網路但有不同的觀測值數量的資料，以分析觀測值數量對演算法表現的影響。另外我們也用微陣列晶片(microarray)資料來重建生物的基因網路：focal adhesion pathway選出的基因網路、SGS1及其synthetic sick or lethal(SSL) partners以及G2/M DNA damage checkpoint的基因網路，用來評估演算法的效能。 就模擬的資料來看，即使是是短時間序列(14個觀測值)，演算法仍然可以保有一定的效能，而長時間序列(52個觀測值)的效能更佳。而就微陣列晶片(microarray)資料來看，靈敏度(sensitivity)或正確率(true positive rate)可以達到50%左右。

With the continual progress of human genome researches, more and more genes have been found to be closely related to human diseases. Accordingly, exploration of genetic functions has become one of major foci in biotechnology researches. It is well known that each gene does not work alone. Instead, it may involve enormous complicated interactions among genes in a biological process. Because of the complexity of physiological and biochemical processes in biology, the relations between the genes and most diseases are not clear currently. Therefore, the ultimate goal of gene networks reconstruction is to analyze the regulatory mechanisms among genes and understand how genes involve in biological processes. Limited by the high cost of microarrays, most biological experiments can not offer a large number of observations for gene network reconstruction. To overcome this limitation, a new gene network model：linear dynamic factor model, which is based on structural equation modeling, is proposed in this study. Besides observed variables, linear dynamic factor model also incorporates hidden factors to depict regulations from proteins and other molecules that are not included in the gene networks but have influence on the gene networks. We simulated data from a 6-gene network with different observations to see the influence of the number of observations on the performance of the algorithm. We also applied the algorithm to microarray data to reconstruct the gene networks from focal adhesion pathway、SGS1 and its synthetic sick or lethal(SSL) partners and G2/M DNA damage checkpoint of Saccharomyces cerevisiae. For the simulated data with 14 observations, the performance of the algorithm is well；for the simulated data with 52 observations, the performance of the algorithm is better than that of the simulated data with 14 observations. For the microarray data, the sensitivity or true positive rate can be in the neighborhood of 50%.