Coupling a multifunctional tag to scalable endogenous tagging technology for improved genome-wide perturbation screens

项目摘要 确定蛋白质编码基因的功能是后基因组时代的一个重要目标。而当 蛋白质是大多数细胞功能的最终效应者，包括那些在疾病中被错误调节的功能，我们有 对大多数蛋白质在人类蛋白质组中的作用的了解极其有限。尽管 用于高通量询问蛋白质编码基因的强大的现有技术，包括 基于CRISPR/Cas9的方法和RNA干扰需要较长的时间才能影响 蛋白质水平，并因此遭受两个严重的缺陷。首先，他们无法察觉增长的贡献-- 基因对除生存能力以外的任何细胞过程都是必不可少的基因，如任何携带这种基因扰动的细胞 将无法传播。第二，补偿性和适应性效应有充分的机会表现出来，因此 通过改善扰动的效果或通过产生新的、无关的 效果。为了解决这些关键限制，我建议开发一种新的筛查技术，它将使 通过诱导和快速降解内源蛋白质的扰动和屏幕读出之间的时间。这 是由一种易于扩展的内源标记技术实现的，该技术利用了与同源无关的 靶向整合，将合成的外显子插入到蛋白质编码基因的内含子中 断线了。合成的外显子将编码一种多功能的配体结合蛋白，这取决于 配基，会导致荧光或快速降解。针对不同内含子的sgRNA池文库允许 用于创建定制的细胞库，其中每个细胞在不同的蛋白质上带有这种多功能标签。 这种方法的实用性将通过测试(1)细胞是否经历了 在短暂的扰动结束时，细胞文库中维持着生长必需蛋白的快速耗尽 窗口和(2)快速耗尽和CRISPR基因敲除在同一蛋白质上是否产生不同的影响 由于淘汰赛中适应事件的扭曲效应，这是一个成熟的表型。AIM 3目击者 使用机器学习方法和来自数千次尝试标记事件的数据来识别 潜在标签位置的特征如何决定携带多功能的功能蛋白的可能性 将生成标签。由此产生的模型将在蛋白质编码基因组上释放，以预测高质量 标记尽可能多的蛋白质编码基因的位置。这将建立一个改进的筛查范例， 将允许集中询问数千种蛋白质对感兴趣的表型的贡献，将 从而加快了我们理解蛋白质编码中鲜为人知的元素的速度 基因组。这些努力将得到实验和指导方面的杰出资源的很好支持 在宾夕法尼亚大学和费城儿童医院，并将提供出色的 关于蛋白质扰动和表征的实验技术以及计算能力的培训 在广泛有用的机器学习领域。

Project Summary Characterizing the functions of protein-coding genes is an important goal in the post-genomic era. While proteins are the ultimate effectors of most cellular functions, including those mis-regulated in disease, we have an extremely limited understanding of the roles of the majority of proteins in the human proteome. Though powerful, existing technologies for the high-throughput interrogation of protein-coding genes, including CRISPR/Cas9-based approaches and RNA interference, require extended periods of time to effect changes in protein levels, and thus suffer two critical shortcomings. First, they are unable to detect the contribution of growth- essential genes to any cellular process other than viability, as any cell carrying a perturbation in such a gene would fail to propagate. Second, compensatory and adaptive effects have ample opportunity to manifest, thus convoluting screen results by ameliorating the effect of the perturbation, or by generating a novel, unrelated effect. To address these critical limitations, I propose to develop a new screening technology that will minimize the time between perturbation and screen readout by inducibly and rapidly degrading endogenous proteins. This is made possible by a readily scalable endogenous tagging technology that harnesses homology-independent targeted integration to insert a synthetic exon into the intron of a protein-coding gene at the site of a double strand break. The synthetic exon will encode a multifunctional ligand-binding protein that depending on the ligand, will lead to fluorescence or rapid degradation. Pooled libraries of sgRNAs targeting different introns allows for the creation of custom libraries of cells, where each cell carries this multifunctional tag on a different protein. The utility of this approach will be established aims 1 and 2 by testing (1) whether cells that have undergone rapid depletion of growth-essential proteins are maintained in the cell library at the end of the short perturbation window and (2) whether rapid depletion and CRISPR knockout at the same protein produce different effects on a well-established phenotype, due to the distorting effects of adaptation events in the knockout. Aim 3 witnesses the use of a machine learning approach and the data from thousands of attempted tagging events to identify how the features of a potential tag site dictate the likelihood that a functional protein carrying the multifunctional tag will be produced. The resulting model will be unleashed on the protein-coding genome to predict high-quality tag sites for as many protein-coding genes as possible. This will establish an improved screening paradigm that will allow for the pooled interrogation of the contributions of thousands of proteins to a phenotype of interest, will thus accelerate the rate at which we come to understand the poorly understood elements of the protein-coding genome. These efforts will be well supported by the outstanding resources for experimentation and mentorship at both the University of Pennsylvania and the Children’s Hospital of Philadelphia, and will provide excellent training in experimental techniques for protein perturbation and characterization, as well computational literacy in the broadly useful field of machine learning.