Wang Zengmiao, Wang Jun, Wu Changjing, Deng Minghua
* Center for Quantitative Biology, Peking University, Beijing 100871, P. R. China.
† School of Mathematical Sciences, Peking University, Beijing 100871, P. R. China.
J Bioinform Comput Biol. 2015 Dec;13(6):1542001. doi: 10.1142/S0219720015420019. Epub 2015 Aug 11.
Estimation of gene or isoform expression is a fundamental step in many transcriptome analysis tasks, such as differential expression analysis, eQTL (or sQTL) studies, and biological network construction. RNA-seq technology enables us to monitor the expression on genome-wide scale at single base pair resolution and offers the possibility of accurately measuring expression at the level of isoform. However, challenges remain because of non-uniform read sampling and the presence of various biases in RNA-seq data. In this paper, we present a novel hierarchical Bayesian method to estimate isoform expression. While most of the existing methods treat gene expression as a by-product, we incorporate it into our model and explicitly describe its relationship with corresponding isoform expression using a Multinomial distribution. In this way, gene and isoform expression are included in a unified framework and it helps us achieve a better performance over other state-of-the-art algorithms for isoform expression estimation. The effectiveness of the proposed method is demonstrated using both simulated data with known ground truth and two real RNA-seq datasets from MAQC project. The codes are available at http://www.math.pku.edu.cn/teachers/dengmh/GIExp/.
基因或异构体表达的估计是许多转录组分析任务中的基本步骤，如差异表达分析、eQTL（或sQTL）研究以及生物网络构建。RNA测序技术使我们能够在全基因组范围内以单碱基对分辨率监测表达，并提供了在异构体水平上准确测量表达的可能性。然而，由于读取采样不均匀以及RNA测序数据中存在各种偏差，挑战依然存在。在本文中，我们提出了一种新颖的分层贝叶斯方法来估计异构体表达。虽然大多数现有方法将基因表达视为副产品，但我们将其纳入模型，并使用多项分布明确描述其与相应异构体表达的关系。通过这种方式，基因和异构体表达被纳入一个统一的框架，这有助于我们在异构体表达估计方面比其他现有算法取得更好的性能。使用具有已知真实情况的模拟数据以及来自MAQC项目的两个真实RNA测序数据集证明了所提出方法的有效性。代码可在http://www.math.pku.edu.cn/teachers/dengmh/GIExp/获取。
