Functional Genomics and Genetic Screens

Introduction to Functional Genomics

The sequencing of the human genome in 2001 provided a parts list—the approximately 20,000 protein-coding genes and millions of regulatory elements. But knowing the sequence does not reveal gene function. Functional genomics uses high-throughput experimental approaches to systematically assign biological functions to genomic elements—through gene perturbation screens (inactivating or modifying each gene separately or simultaneously), gene expression profiling (measuring responses to perturbations), protein-protein interaction maps, and integrative multi-omics analysis. These approaches are transforming biology from descriptive to mechanistic, identifying drug targets, disease gene functions, and the genetic architecture of complex traits at genome scale.

The transition from classical forward genetics (phenotype first → gene mapping) to reverse genetics (gene first → phenotype) and now to genome-scale functional interrogations represents a fundamental shift enabled by CRISPR genome editing tools, arrayed and pooled screening technologies, and single-cell sequencing. Genome-wide CRISPR loss-of-function screens in human cells can now simultaneously test all ~20,000 essential genes for contribution to a phenotype in a single experiment—a task that would have required 20,000 separate experiments with earlier technologies. DepMap (Cancer Dependency Map) has profiled genetic dependencies (CRISPR screens) and drug sensitivities across 1000+ cancer cell lines, creating a public resource linking genomic features to therapeutic vulnerabilities.

CRISPR Functional Screens

Pooled Loss-of-Function Screens

Pooled CRISPR screens deliver a library of tens of thousands of sgRNAs (4-6 per gene) via lentiviral transduction at low MOI (one sgRNA per cell), select for a phenotype over time (viability, drug resistance, reporter activation), then quantify sgRNA representation by deep sequencing comparing pre- and post-selection populations. MAGeCK, BAGEL, and CRISPRscore algorithms statistically identify depleted (essential genes) or enriched (resistance genes) sgRNAs from count comparisons. Genome-wide essential gene screens identified ~2000 core essential genes in cancer cells; context-specific dependencies identify tumour type-specific vulnerabilities (KRAS dependency in KRAS-mutant tumours, AR dependency in prostate cancer, EZH2 in NSD1-mutant tumours). The Broad Institute DepMap project and Sanger Institute Project Score have profiled dependencies across hundreds of cancer cell lines with matched genomic data—identifying genomic biomarkers of specific dependencies.

CRISPRa and CRISPRi Screens

CRISPRa (activation) uses dCas9 fused to transcriptional activators (VP64, VPR, SAM system with MS2 RNA aptamers recruiting additional activators) to upregulate gene expression from endogenous promoters; CRISPRi (interference) uses dCas9 fused to KRAB repressor domain to silence transcription without DNA cutting. Genome-scale CRISPRa screens identify gain-of-function suppressors or synthetic rescue factors; CRISPRi provides more uniform and reversible gene repression than nuclease-based knockouts enabling quantitative phenotypic analysis across gene dosage ranges. CRISPRa in vivo screens in mice identified metabolic regulatory genes suppressing non-alcoholic liver disease; CRISPRi screens in T cells identified checkpoint resistance mechanisms. Base editing screens use cytidine and adenine base editors to introduce disease-associated point mutations across the genome systematically, creating saturation mutagenesis maps informing variant pathogenicity interpretation.

Perturb-seq and Multi-Modal Screens

Single-Cell CRISPR Screens

Perturb-seq (and variants CROP-seq, Mosaic-seq) combines CRISPR perturbations with single-cell transcriptomics—reading out genome-wide expression responses for each perturbation individually in a pooled format. Cells receive individual sgRNAs via lentiviral delivery; sgRNA identity is read from the scRNA-seq library using sgRNA-capturing barcodes expressed within the sgRNA construct. This enables: phenotypic response to each gene KO at the transcriptome level; identifying gene regulatory relationships and gene network logic; classifying perturbation effects into pathways by clustering transcriptional response patterns; and assigning non-overlapping functions to genes within a pathway without confounding by secondary effects. Genome-scale Perturb-seq experiments with up to 2 million cells (Replogle et al. 2022) profiled every essential gene's transcriptional effect simultaneously—creating the first genome-scale perturbation atlas of transcriptional responses.

Multi-Omics Integration

No single omic layer fully explains biological function—integrating multiple data types (genome variants → eQTL → proteomics → metabolomics → phenome) provides mechanistic chains from genetic cause to phenotypic consequence. DepMap integrates CRISPR dependency data with multi-omic cancer cell line profiles (copy number, mutation, methylation, gene expression, proteomics) to build predictive models of therapeutic vulnerability. The MoTrPAC (Molecular Transducers of Physical Activity Consortium) integrated 25 molecular assays across 19 tissues before and after endurance exercise training in rats, connecting exercise-responsive multi-omic nodes to phenotypic health outcomes. Spatial transcriptomics (Visium, Slide-seq, MERFISH) adds tissue spatial dimension to multi-omics integration—mapping gene expression to tissue architecture for understanding cellular communication in its anatomical context.

Examples and Applications

Example 1: Synthetic Lethality Screens

Synthetic lethality—two separate perturbations each individually tolerated but lethal in combination—provides cancer-specific therapeutic windows. PARP inhibitors exploit synthetic lethality in BRCA1/2-mutant cancers: BRCA1/2 loss impairs homologous recombination (HR); PARP inhibition stalls replication forks requiring HR for resolution; BRCA-deficient cells cannot resolve stalled forks and die. Olaparib, rucaparib, niraparib, and talazoparib approvals in BRCA-mutant breast, ovarian, prostate, and pancreatic cancers represent the most successful synthetic lethality clinical translation. Genome-wide CRISPR synthetic lethality screens identify further pairs—WRN helicase dependence in microsatellite-unstable cancers, POLRMT dependence in MYC-amplified cancers, MTHFD2 in many cancer genotypes—each providing potential synthetic lethality-based new drugs targeting tumour-specific vulnerabilities.

Example 2: GWAS Function Assignment

Most GWAS associations (>90%) fall in non-coding intergenic or intronic regions—challenges for interpreting functional mechanism. Functional genomics connects GWAS signals to causal variants and genes: eQTLs (expression QTLs)—variants affecting nearby gene expression in specific tissues from GTEx—identify the likely causal gene when GWAS and eQTL signals colocalize (tested by statistical colocalisation). ENCODE DNase-seq and H3K27ac ChIP-seq data identifies whether GWAS variants fall in active enhancers or promoters in relevant cell types. CRISPR tiling approaches saturate-mutagenise specific GWAS regions identifying functional nucleotides. Fine-mapping (SuSiE, FINEMAP) narrows credible sets from hundreds of candidates to a few variants. Variant-to-function workflows integrating GWAS, QTLs, and functional annotation are now standard analytical frameworks in human genetics for identifying drug targets from common disease association studies.

Example 3: RNA Interference Screens

RNA interference (RNAi) screens using siRNA libraries targeting each human gene were the first genome-scale loss-of-function approaches in human cells (before CRISPR). Genome-wide siRNA screens identified essential genes for viral replication (HIV dependency factors: TSG101, nuclear transport genes, NPC1 for Ebola entry), cancer cell line essentials, and drug target identification. RNAi screens have significant off-target effects (seed sequence cross-reactivity) complicating interpretation; CRISPR KO and CRISPRi provide cleaner perturbations and have largely superseded siRNA for pooled screens. Custom siRNA/ASO screens remain valuable for specific applications (high-content imaging screens for morphological phenotypes; screens in primary cells where lentiviral MOI is too sensitive). shRNA libraries (VIPER, Sigma MISSION) provide stable long-term knockdown for in vivo screens not feasible with transient siRNA transfection.

Example 4: High-Content Imaging Screens

High-content imaging (HCI) screens combine automated confocal microscopy with machine-learning image analysis to measure complex morphological phenotypes in genome-wide perturbation screens. Cell Painting—a 6-fluorophore staining protocol marking nucleus, cytoplasm, mitochondria, ER, nucleoli, and F-actin simultaneously—generates ~1500 morphological features per cell for phenotypic fingerprinting. Morphological profiles from CRISPR knockouts or compound treatments cluster by shared biology—genes in the same pathway show similar morphological profiles enabling pathway discovery. Rxrx (Recursion Pharmaceuticals) Cell Painting screens of millions of compound-cell combinations generated the largest morphological imaging dataset in drug discovery. Human Cell Map project systematically profiles the spatial localisation of human proteome using HCI fluorescence labelling—creating an image-based atlas of protein localisation comparable to the Human Protein Atlas.

Example 5: CRISPR Knockin for Saturation Genome Editing

Saturation genome editing (SGE) introduces every possible single nucleotide variant (SNV) in a target region—typically encompassing a clinical gene or locus of interest—and measures each variant's effect on cell fitness or a specific readout. Findlay et al. (2019) SGE of BRCA1 exons tested 3,893 single-nucleotide variants simultaneously, functionally classifying each as functional, intermediate, or non-functional—creating a clinical interpretation ground truth for >95% of BRCA1 exonic variants months before they are ever encountered clinically. SGE enables pre-emptive clinical variant classification at scale rather than waiting for uncertain variants to be encountered in patients. Similar approaches are being applied to TP53, PTEN, MSH2, LDLR, and other clinically important genes—generating empirical functional classification for all possible variants as a clinical genomics resource.

Example 6: DNA Damage Response Screens

Ionising radiation and chemotherapy screens identify genes whose loss sensitises or confers resistance to DNA-damaging agents, informing combination therapy rationales and predictive biomarkers. Genome-wide CRISPR screens in cancer cells treated with PARP inhibitors, cisplatin, or olaparib identified genes including PARG (poly-ADP ribose glycohydrolase—its inhibition sensitises PARP inhibitor-resistant cells), RNF8/168 (ubiquitin ligases in HR pathway), and 53BP1/SHIELDIN complex (which antagonises HR—its loss restores HR in BRCA1/2-mutant cancers, conferring PARP inhibitor resistance but olaparib AUC improvement after chemotherapy). CTC1-STN1-TEN1 complex loss confers sensitivity to ATR inhibitors explaining ATR inhibitor synthetic lethality in CDC25A-expressing cancers. These mechanistic insights from genome-scale screens directly inform clinical trial design for combination therapies and patient stratification.

Example 7: In Vivo CRISPR Screens

CRISPR screens in animal models identify genes operating in physiological contexts inaccessible to cell culture. In vivo CRISPR screens in mouse models of tumourigenesis identify metastasis suppressor and promoter genes by comparing sgRNA abundance in primary tumours versus metastases—identifying ETV6, PARD3 loss promoting breast cancer liver metastasis. In vivo screens in CAR-T cells identified REGNASE-1 as a negative regulator—its deletion improved CAR-T anti-tumour function in solid tumours. Non-alcoholic fatty liver disease screen using AAV-CRISPR to perform liver-specific KO screens in mice on high-fat diet identified NSD3 and previously unknown metabolic regulators. Pooled AAV delivery enables liver, lung, and CNS in vivo screens in mice; future in vivo screens in non-human primates or patient-derived xenograft models would further bridge preclinical-clinical translation.

Example 8: Base Editing Screens for Clinical Variants

Base editing screens systematically introduce point mutations in disease-relevant loci and measure functional consequences at population scale. Large-scale adenine base editing (ABE) screens in CML cells tiling BCR-ABL kinase domain introduced thousands of single amino acid variants measuring resistance to imatinib, dasatinib, and asciminib—generating resistance mutation maps predicting which not-yet-observed clinical resistance mutations will emerge under each drug. Cytidine base editing (CBE) screens in EGFR-mutant lung cancer cells predicted acquired resistance to osimertinib. Integrating these resistance mutation maps into clinical pharmacology and combination therapy design could pre-empt resistance before it develops clinically. Base editing library screens with PGT-competent cells could validate candidate therapeutic base edits for cystic fibrosis, sickle cell disease, and other Mendelian diseases for advancing to clinical gene therapy applications.

Try it live

Everything above runs in your browser — open CRISPR Genetic Screen Simulator and change the parameters while it is running. Nothing is installed, nothing is uploaded, the whole model lives in one tab.