Yang E, Almon R R, Dubois D C, Jusko W J, Androulakis I P
Department of Biomedical Engineering, Rutgers University; 599 Taylor Rd., Piscataway, NJ 08544, USA.
J Pharmacol Exp Ther. 2008 Mar;324(3):1243-54. doi: 10.1124/jpet.107.133074. Epub 2007 Dec 17.
One of the challenges in constructing biological models involves resolving meaningful data patterns from which the mathematical models will be generated. For models that describe the change of mRNA in response to drug administration, questions exist whether the correct genes have been selected given the myriad transcriptional effects that may occur. Oftentimes, different algorithms will select or cluster different groups of genes from the same data set. A new approach was developed that focuses on identifying the underlying global dynamics of the system instead of selecting individual genes. The procedure was applied to microarray genomic data obtained from rat liver after a large single dose of methylprednisolone in 52 adrenalectomized rats. Twelve clusters of at least 30 genes each were selected, reflecting the major changes over time. This method along with isolating the underlying dynamics of the system also extracts and clusters the genes that make up this global dynamic for further analysis as to the contributions of specific mechanisms affected by the drug.
构建生物学模型面临的挑战之一是从有意义的数据模式中解析出数学模型。对于描述mRNA响应药物给药变化的模型，鉴于可能发生的无数转录效应，是否选择了正确的基因仍存在疑问。通常，不同的算法会从同一数据集中选择或聚类不同的基因组。开发了一种新方法，该方法侧重于识别系统的潜在全局动态，而不是选择单个基因。该程序应用于52只肾上腺切除大鼠单次大剂量注射甲基强的松龙后从大鼠肝脏获得的微阵列基因组数据。选择了12个每组至少30个基因的聚类，反映了随时间的主要变化。这种方法除了分离系统的潜在动态外，还提取并聚类构成这种全局动态的基因，以便进一步分析受药物影响的特定机制的贡献。
