Computational tools for RNA sequencing power analysis and data integration

RNA测序为了解细胞的运作提供了一个窗口，可以量化几乎所有表达基因的水平。不同细胞之间基因表达的差异以及对不同环境的反应可以提供生物化学途径和细胞功能的重要信息。然而，能够有效地研究差异基因表达依赖于具有准确和适当的计算工具来解释数据。这项拨款申请涉及基因表达数据分析的两个重要方面——(i)可以检测到基因表达变化的大小，即实验的力量是什么；（ii）不同实验中的基因反应有多相似，是否可以准确确定？在发现基金内进行的研究将通过使用重采样技术来估计RNA测序实验的事后能力，包括实验范围和每个基因（目标1）。这将为任何高通量表达实验提供重要的数据，因为将估计所有基因（差异表达与否）的可检测效应的大小。其次，我们使用一种基于统计对比的方法来比较不同实验的数据（目标2）。我们不是比较由单独实验产生的差异表达基因列表，而是确定(i)在两个实验中基因表达变化相似的基因，以及在两个实验之间差异表达的基因。事后看来，这种方法似乎清晰有效，但在文献中缺乏类似方法的应用实例。最后，我们将通过探索针对细胞系中特定生化途径的影响来验证我们的方法（Aim 3），扩展McConkey实验室之前的工作。这项工作将为更可靠和深入地探索RNA测序表达数据集创造工具。重要的是，它将允许研究人员更好地量化可以检测到的基因表达变化的幅度（功率分析），而不是仅仅关注被识别为差异表达的内容（显著性检验）。其次，在多个实验之间进行比较的更好的工具将极大地有利于研究人员使用高通量表达研究，以快速识别可能对给定扰动具有高度特异性的基因，而不是“常见应答者”，如代谢或应激反应基因。

RNA sequencing provides a window into the operation of the cell and can quantify levels of almost all expressed genes. Differences in gene expression between different cells and in response to different environments can provide important information on biochemical pathways and how the cell is functioning. Being able to effectively investigate differential gene expression however relies on having accurate and appropriate computational tools to interpret the data. This grant application addresses two important aspects of data analysis of gene expression - (i) what size of change in gene expression can be detected, i.e. what is the power of the experiment; and (ii) how similar are the gene responses across different experiments, and can these be determined accurately? The research conducted within the discovery grant will address this by using resampling techniques to estimate post hoc power of RNA sequencing experiments, both the experiment-wide level and for each gene (Aim 1). This will provide important data for any high-throughput expression experiment, as the magnitude of detectable effect will be estimated for all genes (differentially expressed or not). Secondly, we are using a method based on statistical contrasts to compare data across experiments (Aim 2). Instead of comparing lists of differentially expressed genes produced by separate experiments, we identify (i) genes that have similar changes in gene expression in both experiments, as well as genes that are differentially expressed between experiments. In hindsight this method seems clear and effective, yet there is a lack of examples in the literature of applications of similar methodology. Lastly, we will verify our methods by exploring the effects of targeting specific biochemical pathways in cell lines (Aim 3), extending previous work in the McConkey lab. This work will create tools for more reliable and in-depth exploration of RNA sequencing expression data sets. Importantly, it will permit researchers to better quantify the magnitude of changes in gene expression that can be detected (power analysis) instead of focusing exclusively on what is identified as differentially expressed (significance testing). Secondly, better tools to compare across multiple experiments will be hugely beneficial to researchers using high-throughput expression studies, to quickly identify genes that may have a high degree of specificity to a given perturbation, versus `common responders' such as metabolic or stress response genes.