Scalable, compressed phenotypic screening using pooled perturbations (PDAC screens). Scalable, compressed phenotypic screening using pooled perturbations (PDAC screens)

High-throughput phenotypic screens leveraging biochemical perturbations and high-content readouts are poised to advance therapeutic discovery, yet they remain constrained by limitations of scale. To address this, we establish a method of pooling exogenous perturbations followed by computational deconvolution to compress a screen’s required sample, labor, and financial input. We benchmark the approach with a bioactive small molecule library and a high-content imaging readout, demonstrating the feasibility and increased efficiency of compressed experimental designs compared to conventional approaches. To prove generalizability, we apply compressed screening in two different biological discovery campaigns. In the first, we use early-passage pancreatic cancer organoids to map transcriptional responses to alibrary oftumor-microenvironmentrecombinant protein ligands. We uncover reproducible phenotypic shifts induced by specific ligands that are distinct from canonical reference signatures and uniquely correlate with clinical outcome.In the second, we examine the modulatory effects of a known mechanism of action chemical compound library on primary human peripheral blood mononuclear cell immune responses. Through contrastive analyses, we identify molecules that potentiate and/or inhibit cell-type specific transcriptional features, uncover pleiotropic effects for individual compounds across diverse cell types, and realize a systems-level view of drug responses. In sum, our approach empowers phenotypic screens with information-rich readouts to advance drug discovery efforts as well as basic biological inquiry. Overall design: Patient-derived pancreatic cancer (PDAC) organoids from patients with metastatic pancreatic cancer (Raghavan et al., Cell, 2021) were screened with a library of 68 macrophage-derived ligands (and DMSO control) for the compressed screen and 11 ligands (and DMSO control) for the validation experiment for 7 days and analyzed using scRNA-seq.

借助生化扰动与高内涵检测（high-content readouts）的高通量表型筛选（high-throughput phenotypic screens），有望助力治疗性药物研发，但此类筛选目前仍受限于规模瓶颈。为解决这一问题，本研究提出一种将外源性扰动合并后辅以计算反卷积（computational deconvolution）的方法，可压缩筛选所需的样本量、人力与资金投入。本研究使用生物活性小分子文库与高内涵成像检测（high-content imaging readout）对该方法进行基准验证，证明相较于传统实验方案，压缩式实验设计具备可行性且效率更高。为验证方法的普适性，我们在两项独立的生物学发现研究中应用压缩筛选技术。第一项研究中，我们使用早期传代胰腺癌类器官，绘制肿瘤微环境重组蛋白配体文库刺激下的转录响应图谱；我们发现特定配体可诱导出可重复的表型变化，此类变化不同于经典参考特征集，且与临床结局存在特异性关联。第二项研究中，我们针对已知作用机制的化合物文库，探究其对原代人外周血单个核细胞免疫响应的调控作用。通过对比分析，我们筛选出可增强或抑制细胞类型特异性转录特征的分子，揭示了单一化合物在不同细胞类型中的多效性作用，并构建了药物响应的系统级视图。综上，本方法可为搭载高信息量检测读数的表型筛选提供技术赋能，助力药物研发与基础生物学研究。整体实验设计：本研究使用来自转移性胰腺癌患者的患者源性胰腺癌（Patient-derived pancreatic cancer, PDAC）类器官（参考Raghavan等发表于《Cell》2021年的研究），压缩筛选实验采用包含68种巨噬细胞源性配体与二甲基亚砜（dimethyl sulfoxide, DMSO）对照的文库，验证实验则采用包含11种配体与DMSO对照的文库，所有样本均培养7天后通过单细胞RNA测序（single-cell RNA sequencing, scRNA-seq）进行分析。