Predicting SNP priors for Cancer GWASs

摘要 目前用于GWASs的基因分型平台包括数百万个SNP。一大批人的基因分型 SNPs是有益的，因为有更高的机会直接对因果变异进行基因分型。霍沃尔， 对多个SNP进行基因分型会增加由于多次检测而被错误发现的机会。 识别癌症风险易感性的基因控制是一个热门的研究领域 十年。已经进行了数百次GWAS，许多与风险相关的SNPs 已确认身份。现在是时候学习已经发表的癌症GWAS来建立一个更高效的 确定与癌症风险相关的SNP的有针对性的战略。 基于我们和其他人的数据，我们提出了三个假设：(I)癌症风险相关的SNPs 共有一组共同特征；(2)这些特征可用于识别具有高基因频率的SNPs 与癌症风险有关的倾向；(Iii)合并现有的几个GWASs 对癌症风险基因预测因子的了解可能会大大提高识别的统计能力 新的癌症相关SNPs。我们将对现有的基因分型数据进行有针对性的分析 五种最常见的癌症：乳腺癌、结肠癌、肺癌、卵巢癌和前列腺癌。我们的初步结果表明 我们将能够识别出对癌症风险影响较小的SNPs。我们将使用DBGaP数据来 用于验证的Discovery和OncoArray数据。我们提出了三个具体目标： 目的1.建立预测SNPs概率的癌症类型特异性模型和泛癌模型 与癌症风险有关。 目的2.对预测致癌概率最高的灰色地带SNPs进行靶向分析 相关风险。 目标3.使用独立样本对目标2中最重要的单核苷酸多态进行验证。 该项目的主要目标是确定具有最高癌症风险概率的SNPs 并使用它们在针对五种最常见癌症的硅胶GWAS中进行靶向治疗。我们 将确定与已检测到的风险关联SNP相结合的新风险关联SNP 将更好地解释癌症易感性的基因控制。拟议的方法代表了以下成本： 传统GWAS的有效替代品。

ABSTRACT Current genotyping platforms for GWASs include millions SNPs. Genotyping of a large number of SNPs is beneficial because of higher chances of direct genotyping of causal variant. Hovewer, genotyping multiple SNPs elevates chances of false discoveries because of multiple testing. Identification of genetic control of cancer risk susceptibility was a hot research area for more than decade. Several hundred GWASs have been conducted and many risk associated SNPs were identified. Now it is time to learn from already published cancer GWASs to build a more efficient targeted strategy to identify cancer risk associated SNPs. Based on ours and others data we propose three-fold hypothesis: (i) Cancer risk associated SNPs share sets of common characteristics; (ii) Those characteristics can be used to identify SNPs with high propensity to be associated with cancer risk; (iii) Combining several GWASs enriched by existing knowledge on the cancer-risk gene predictors may substantially improve statistical power to identify novel cancer-associated SNPs. We will conduct targeted analyses of existing genotyping data for the five most common cancers: breast, colon, lung, ovarian, and prostate. Our preliminary results indicate that we will be able to identify SNPs with smaller effects on cancer risk. We will use dbGAP data for discovery and OncoArray data for validation. We propose 3 specific aims: Aim 1. To develop cancer type specific as well as pan-cancer models for predicting SNPs' probabilities to be associated with cancer risk. Aim 2. To conduct targeted analysis of gray zone SNPs with highest predicted probability to be cancer risk associated. Aim 3. To perform validation of the top most significant SNPs from Aim 2 using independent samples. Primary goal of this project is to identify SNPs with highest prior probabilities to be cancer risk associated and use them to perform targeted in silico GWASs for the five most common cancers. We will identify novel risk associated SNPs that in combination with already detected risk associated SNPs will better explain genetic control of cancer susceptibility. The proposed approach represents a cost- effective alternative of traditional GWASs.