A scalable image-based approach for profiling and annotating very large compound

描述（由申请人提供）：迫切需要大幅增加可用于对抗癌症的药物库：许多药物具有严重的副作用，并且耐药性可能迅速出现。使我们目前选择的化疗剂多样化的一种有效方法是鉴定在细胞中显示出与已证实的药物相似的作用但靶向不同途径组分或作用机制的化合物。 大型化合物库越来越多地可用，并作为此类搜索的起点。然而，这些库中的绝大多数化学品的生物活性是完全未知的。目前，整个化合物库需要重新筛选，以完成看似简单的任务，即查询具有相似生物功能的更多药物，这是一个昂贵，耗时且效率低下的过程。 高维表型筛选非常适合表征系统水平对跨多个途径和遗传背景的化合物的响应。然而，转录组学或蛋白质组学等方法过于昂贵和耗时，无法常规地扩展到具有数十万种化合物的文库。 一个强大的方法来注释化合物库与预测的生物功能是使用显微镜为基础的细胞学配置文件，一种方法来量化细胞的扰动反应，我们的实验室在过去的十年中开创了。通常，仅需要一个屏幕来获得可用于预测跨多个功能类别的作用机制的配置文件。虽然这种方法显示出前景，但由于试剂的高成本和关于哪种细胞读数最好地允许区分广泛类别的生物功能的不确定性，其使用仅限于小型化合物库。 在目标1中，我们克服了目前的局限性，开发了一种可扩展且具有成本效益的方法，用于识别对多种已证实的癌症药物具有类似反应的化合物。在目标2中，我们在一个中等规模的化合物库上校准和测试我们的方法。在目标3中，我们注释了包含数十万种化学物质的大型化合物库，在多个已证实的药物类别中识别出高价值的治疗前先导化合物，并搜索具有全新作用机制的化合物。我们的目标将为加快癌症药物发现的步伐提供新的范例。

DESCRIPTION (provided by applicant): There is a pressing need to dramatically increase the repertoire of drugs available to fight cancer: many drugs have severe side effects and drug resistance can rapidly emerge. An effective approach for diversifying our current selection of chemotherapeutic agents is to identify compounds that show similar effects to proven drugs in cells, but target different pathway components or mechanisms of action. Large compound libraries are becoming increasingly available and serve as starting points for such searches. However, biological activities are completely unknown for the vast majority of chemicals in these libraries. Currently, entire compound libraries need to be re-screened for the seemingly simple task of querying for more drugs with similar biological function-a process that is costly, time consuming and inefficient. High-dimensional phenotypic screens are well suited to characterize systems-level responses to compounds across multiple pathways and genetic backgrounds. However, approaches such as transcriptomics or proteomics are far too expensive and time consuming to be scaled routinely to libraries with hundreds of thousands of compounds. A powerful method for annotating compound libraries with predicted biological function is the use of microscopy-based cytological profiles, an approach for quantifying cellular responses to perturbations that our lab has pioneered over the past decade. Often, only a single screen is necessary to obtain profiles that can be used to predict mechanisms of action across multiple functional categories. Although this approach shows promise, its use has been limited to small compound libraries due to the high cost of reagents and uncertainty about which cellular readouts would best allow broad classes of biological function to be distinguished. In Aim 1, we overcome current limitations and develop a scalable and cost-effective approach for identifying compounds that give similar responses to multiple classes of proven cancer drugs. In Aim 2, we calibrate and test our approach on a medium-sized compound library. In Aim 3, we annotate large compound libraries containing hundreds of thousands of chemicals, identify high-value pre-therapeutic leads in multiple proven drug categories, and search for compounds with completely novel mechanisms of action. Our Aims will provide a new paradigm for accelerating the pace of cancer drug discovery.