Abstract 3550: PerturbDB: A resource for revealing gene functions and cancer-associated phenotypes

Abstract Perturb-seq is a high-throughput technique that combines clustered regularly interspaced short palindromic repeats (CRISPR)-based screens with single-cell RNA sequencing (scRNA-seq) readouts for high-content phenotypic screens to comprehensively map the transcriptional effects of genetic perturbations, showing great advantages in revealing disease-associated gene functions and mechanisms in high volumes. Despite the rapid accumulation of Perturb-seq datasets, no dedicated database exists for reusing these valuable information. In this study, we developed a platform called PerturbDB (http://research.gzsys.org.cn/perturbdb) to facilitate users to unveil genotype-phenotype relations, especially gene functions and regulatory networks involved in several classical cancer associated phenotypes using 37 Perturb-seq datasets from 15 studies. By reanalyzing 3,429,829 single-cell transcriptomes from the knockdown of 3214 genes across 10 different cell lines, we identified 749 classical cancer phenotype-related genes and 373 functional gene clusters annotating potential novel functions. Utilizing Marker genes scoring and the InferCNV algorithm, we identified genes involved in 9 malignant phenotypes, which consists of sustaining proliferative signaling (362 genes), resisting cell death (120 genes), inducing angiogenesis (145 genes), tissue invasion and metastasis (321 genes), enabling replicative immortality (160 genes), deregulating cellular metabolism (247 genes), avoiding immune destruction (95 genes), unlocking phenotypic plasticity (231 genes), and chromosome instability (231 genes). Additionally, functional clusters were calculated utilizing Principal Component Analysis (PCAs) and the HDBSCAN package to introduce unknown gene functions related to the cancer associated phenotypes. Perturbation clusters (PCs) were identified by comparing similarities in the transcriptomes of perturbed cells, suggesting individual genes in the same functional cluster perform similar gene functions, and therefore novel functions of 735 genes can be inferred from the known functions of neighboring genes. The PGSEA tool was developed to facilitate functional analysis of genes not yet included, which helps users to predict gene functions through a combination of PerturbDB datasets and personalized RNA-seq datasets. This study will greatly expand our understanding of genes involved in emblematic cancer associated phenotypes and their coordinated functions and regulatory networks. Citation Format: Bing Yang, Man Zhang, Yanmei Shi, Yi-Ming Dong, Xingyu Ma, Jingyuan Zhang, Daning Lu, Jian-You Liao, Dong Yin. PerturbDB: A resource for revealing gene functions and cancer-associated phenotypes [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 3550.