Dzneladze Irakli, Woolley John F, Rossell Carla, Han Youqi, Rashid Ayesha, Jain Michael, Reimand Jüri, Minden Mark D, Salmena Leonardo
Princess Margaret Cancer Centre, University Health Network, Toronto, Canada.
Department of Medical Biophysics, University of Toronto, Toronto, Canada.
PLoS One. 2018 Feb 7;13(2):e0191510. doi: 10.1371/journal.pone.0191510. eCollection 2018.
Our previous studies demonstrated that INPP4B, a member of the PI3K/Akt signaling pathway, is overexpressed in a subset of AML patients and is associated with lower response to chemotherapy and shorter survival. INPP4B expression analysis in AML revealed a right skewed frequency distribution with 25% of patients expressing significantly higher levels than the majority. The 75% low/25% high cut-off revealed the prognostic power of INPP4B expression status in AML, which would not have been apparent with a standard median cut-off approach. Our identification of a clinically relevant non-median cut-off for INPP4B indicated a need for a generalizable non-median dichotomization approach to optimally study clinically relevant genes. To address this need, we developed Subgroup Identifier (SubID), a tool which examines the relationship between a continuous variable (e.g. gene expression), and a test parameter (e.g. CoxPH or Fisher's exact P values). In our study, Fisher's exact SubID was used to reveal EVI1 as a transcriptional regulator of INPP4B in AML; a finding which was validated in vitro. Next, we used CoxPH SubID to conduct a pan-cancer analysis of INPP4B's prognostic significance. Our analysis revealed that INPP4Blow is associated with shorter survival in kidney clear cell, liver hepatocellular, and bladder urothelial carcinomas. Conversely, INPP4Blow was shown to be associated with increased survival in pancreatic adenocarcinoma in three independent datasets. Overall, our study describes the development and application of a novel subgroup identification tool used to identify prognostically significant rare subgroups based upon gene expression, and for investigating the association between a gene with skewed frequency distribution and potentially important upstream and downstream genes that relate to the index gene.
我们之前的研究表明，INPP4B作为PI3K/Akt信号通路的一员，在一部分急性髓系白血病（AML）患者中过表达，且与化疗反应较低和生存期较短相关。AML中INPP4B表达分析显示频率分布呈右偏态，25%的患者表达水平显著高于大多数患者。75%低/25%高的临界值揭示了INPP4B表达状态在AML中的预后价值，而采用标准的中位数临界值方法则无法体现这一点。我们确定了INPP4B具有临床相关性的非中位数临界值，这表明需要一种可推广的非中位数二分法来优化研究具有临床相关性的基因。为满足这一需求，我们开发了亚组识别工具（SubID），该工具可检测连续变量（如基因表达）与测试参数（如CoxPH或Fisher精确P值）之间的关系。在我们的研究中，使用Fisher精确SubID揭示了EVI1是AML中INPP4B的转录调节因子；这一发现已在体外得到验证。接下来，我们使用CoxPH SubID对INPP4B的预后意义进行泛癌分析。我们的分析表明，INPP4B低表达与肾透明细胞癌、肝肝细胞癌和膀胱尿路上皮癌的生存期较短相关。相反，在三个独立数据集中，INPP4B低表达与胰腺腺癌的生存期延长相关。总体而言，我们的研究描述了一种新型亚组识别工具的开发和应用，该工具用于基于基因表达识别具有预后意义的罕见亚组，并研究频率分布呈偏态的基因与可能与索引基因相关的重要上游和下游基因之间的关联。
