Several pooled libraries for KO screen are available from browsing source Addgene, e.g., GeCKO, H1/H2, Brunello and TKO CRISPR-KO libraries [21,22,23,24]. These libraries contain over 18,000 genes with 4–10 single guide RNAs (sgRNA) per gene. Similarly, CRISPRa and SAM libraries for activation screen and CRISPRi libraries for repression screen are also shared [25,26,27]. Custom libraries are useful for specific studies of interest [28].

The first step in pooled screen using the CRISPR/Cas9 system is to generate a library of perturbed cells with lentiviral infection of an sgRNA library. Viruses are produced by transfecting an sgRNA library into appropriate host cells, e.g., HEK 293FT cells with superior virus production capacity. To avoid confusion in interpretation in case the host cells take up multiple sgRNAs and target multiple genes per cell, low (~0.3) multiplicity of infection (MOI) is ensured by empirically determining the viral titer [27,29,30]. It is important to note the limitation in this step—some models, such as NCI-H660 and VCAP cells of PCa, make it quite challenging to implement tools through lentiviral approaches.

One of the most basic experiments to conduct is to identify genes that impact cell fitness. Since perturbations that decrease the cell fitness will be either depleted or completely absent by the end of the screen, this type of screen is termed a negative selection screen. Negative selection screens are most commonly performed in the field of cancer biology to identify dependencies of tumor cells due to specific mutations, copy number alterations, expression patterns and other targets [24,31,32,33,34]. One of the simplest forms of negative selection screen is to continuously culture cells for extended periods of time in order to identify genes required for cell growth. Such screens have been used to identify both the essential genes required for the cell line tested and a small set of genes that are gene-dependent in a particular cancer cell line [35,36,37]. Another negative selection screen form, performed in cell lines with a given genetic background, is the basis for identifying synthetic lethal interactions, in which simultaneous inhibition of two genes impairs cell viability [38]. The discovery of synthetic interactions will enable targeted therapy of cancer cells, making drugs work only on cells with specific alterations, and can offer a new approach to cancer treatment.

The alternative to the negative selection screen is the positive selection screen, which focuses on cells that have been enriched in the course of time. These screens have been used to identify perturbations which confer resistance to small molecules [14,17,39], conditions [40] and pathogen infections [41,42,43,44,45]. In a positive selection screen, most of the population is eliminated, and the few surviving perturbations may become over 100-fold enriched. Eventually, a single screen can result in both positively and negatively selected phenotypes. For example, a viability-based KO screen of a cancer cell line can reveal both oncogene depletion and enrichment of tumor suppressors. Typically, an intermediate dose of a small molecule can identify both sensitization genes and resistance genes [46].

The marker selection screen aims to identify genetic elements which affect the expression of a particular reporter molecule, and the phenotype is not based on cell viability but on mutations that impact the expression of the marker protein. In this type of screen, the reporter can be genetically engineered by replacing the coding sequence of the gene of interest with the fluorescent marker. Ultimately, fluorescence-activated cell sorting (FACS) allows for the identification of upstream expression regulators by sorting cells with sgRNAs targeting genes which affect the expression of the marker [27,47].

Following the selection step, DNA is collected from surviving cells or FACS-selected cells, and PCR is employed to isolate genomic DNA from the cell population and read the genes responsible for the phenotype. Subsequently, large-scale parallel sequencing is performed using next-generation sequencing (NGS) to cover the regions coding for sgRNAs. Several existing algorithms such as Model-based Analysis of Genome-wide CRISPR/Cas9 KO (MAGeCK) [48,49,50], edgeR [51], Bayesian Analysis of Gene EssentiaLity (BAGEL) [52], CRISPR AnalyzeR for Pooled Screens (caRpools) [53], Platform-independent Analysis of Pooled Screens using Python (PinAPL-Py) [54] and DrugZ [55] can then be used to determine the candidate genes responsible for the observed phenotypes through examining the differences in sgRNA abundance between control and phenotypic samples.

Analysis of the screen provides a ranked list of candidate genes causing the phenotype. In order to assess which genes contribute to the phenotype and how much, validation is essential. The most critical validation method is to assess whether the selected phenotype is indeed reproducible by introducing sgRNAs targeting the gene of interest. Recent improvements in the specificity of sgRNAs have relieved the need to confirm binding to the target, as long as multiple sgRNAs directed to the same genetic element elicit the phenotype. However, if required, analyses such as genomic PCR, RT-qPCR and Western blotting can be performed to evaluate the functional modification of the targeted gene [25,27]. It is important to note that the confirmation of on-target activity does not exclude the possibility of phenotyping due to off-target effects. Hence, continuous validation experiments of phenotypes using not only one sgRNA with the highest on-target activity but also multiple sgRNAs for each gene are critical. Rescue experiments are another way to confirm whether the genetic entity confers the phenotype. The goal is to confirm whether restoring candidate expression to physiological levels in CRISPR/Cas9-edited cells returns the cells to their wild-type state [56].