CoRe: a robustly benchmarked R package for identifying core-fitness genes in genome-wide pooled CRISPR-Cas9 screens

A benchmark of computational methods for correcting biases of established and unknown origin in CRISPR-Cas9 screening data

BACKGROUND

CRISPR-Cas9 dropout screens are formidable tools for investigating biology with unprecedented precision and scale. However, biases in data lead to potential confounding effects on interpretation and compromise overall quality. The activity of Cas9 is influenced by structural features of the target site, including copy number amplifications (CN bias). More worryingly, proximal targeted loci tend to generate similar gene-independent responses to CRISPR-Cas9 targeting (proximity bias), possibly due to Cas9-induced whole chromosome-arm truncations or other genomic structural features and different chromatin accessibility levels.

RESULTS

We benchmarked eight computational methods, rigorously evaluating their ability to reduce both CN and proximity bias in the two largest publicly available cell-line-based CRISPR-Cas9 screens to date. We also evaluated the capability of each method to preserve data quality and heterogeneity by assessing the extent to which the processed data allows accurate detection of true positive essential genes, established oncogenetic addictions, and known/novel biomarkers of cancer dependency. Our analysis sheds light on the ability of each method to correct biases under different scenarios. AC-Chronos outperforms other methods in correcting both CN and proximity biases when jointly processing multiple screens of models with available CN information, whereas CRISPRcleanR is the top performing method for individual screens or when CN information is not available. In addition, Chronos and AC-Chronos yield a final dataset better able to recapitulate known sets of essential and non-essential genes.

CONCLUSIONS

Overall, our investigation provides guidance for the selection of the most appropriate bias-correction method, based on its strengths, weaknesses and experimental settings.

Exome-wide evidence of compound heterozygous effects across common phenotypes in the UK Biobank

The phenotypic impact of compound heterozygous (CH) variation has not been investigated at the population scale. We phased rare variants (MAF ∼0.001%) in the UK Biobank (UKBB) exome-sequencing data to characterize recessive effects in 175,587 individuals across 311 common diseases. A total of 6.5% of individuals carry putatively damaging CH variants, 90% of which are only identifiable upon phasing rare variants (MAF < 0.38%). We identify six recessive gene-trait associations (p < 1.68 × 10-7) after accounting for relatedness, polygenicity, nearby common variants, and rare variant burden. Of these, just one is discovered when considering homozygosity alone. Using longitudinal health records, we additionally identify and replicate a novel association between bi-allelic variation in ATP2C2 and an earlier age at onset of chronic obstructive pulmonary disease (COPD) (p < 3.58 × 10-8). Genetic phase contributes to disease risk for gene-trait pairs: ATP2C2-COPD (p = 0.000238), FLG-asthma (p = 0.00205), and USH2A-visual impairment (p = 0.0084). We demonstrate the power of phasing large-scale genetic cohorts to discover phenome-wide consequences of compound heterozygosity.

CRISPR-Cas9 screens reveal common essential miRNAs in human cancer cell lines

BACKGROUND

Genome-wide functional screening using the CRISPR-Cas9 system is a powerful tool to uncover tumor-specific and common genetic dependencies across cancer cell lines. Current CRISPR-Cas9 knockout libraries, however, primarily target protein-coding genes. This limits functional genomics-based investigations of miRNA function.

METHODS

We designed a novel CRISPR-Cas9 knockout library (lentiG-miR) of 8107 distinct sgRNAs targeting a total of 1769 human miRNAs and benchmarked its single guide RNA (sgRNA) composition, predicted on- and off-target activity, and screening performance against previous libraries. Using a total of 45 human cancer cell lines, representing 16 different tumor entities, we performed negative selection screens to identify miRNA fitness genes. Fitness miRNAs in each cell line were scored using a combination of supervised and unsupervised essentiality classifiers. Common essential miRNAs across distinct cancer cell lines were determined using the 90th percentile method. For subsequent validation, we performed knockout experiments for selected common essential miRNAs in distinct cancer cell lines and gene expression profiling.

RESULTS

We found significantly lower off-target activity for protein-coding genes and a higher miRNA gene coverage for lentiG-miR as compared to previously described miRNA-targeting libraries, while preserving high on-target activity. A minor fraction of miRNAs displayed robust depletion of targeting sgRNAs, and we observed a high level of consistency between redundant sgRNAs targeting the same miRNA gene. Across 45 human cancer cell lines, only 217 (12%) of all targeted human miRNAs scored as a fitness gene in at least one model, and fitness effects for most miRNAs were confined to small subsets of cell lines. In contrast, we identified 49 common essential miRNAs with a homogenous fitness profile across the vast majority of all cell lines. Transcriptional profiling verified highly consistent gene expression changes in response to knockout of individual common essential miRNAs across a diverse set of cancer cell lines.

CONCLUSIONS

Our study presents a miRNA-targeting CRISPR-Cas9 knockout library with high gene coverage and optimized on- and off-target activities. Taking advantage of the lentiG-miR library, we define a catalogue of miRNA fitness genes in human cancer cell lines, providing the foundation for further investigation of miRNAs in human cancer.

A comprehensive clinically informed map of dependencies in cancer cells and framework for target prioritization

Genetic screens in cancer cell lines inform gene function and drug discovery. More comprehensive screen datasets with multi-omics data are needed to enhance opportunities to functionally map genetic vulnerabilities. Here, we construct a second-generation map of cancer dependencies by annotating 930 cancer cell lines with multi-omic data and analyze relationships between molecular markers and cancer dependencies derived from CRISPR-Cas9 screens. We identify dependency-associated gene expression markers beyond driver genes, and observe many gene addiction relationships driven by gain of function rather than synthetic lethal effects. By combining clinically informed dependency-marker associations with protein-protein interaction networks, we identify 370 anti-cancer priority targets for 27 cancer types, many of which have network-based evidence of a functional link with a marker in a cancer type. Mapping these targets to sequenced tumor cohorts identifies tractable targets in different cancer types. This target prioritization map enhances understanding of gene dependencies and identifies candidate anti-cancer targets for drug development.

Inferring compound heterozygosity from large-scale exome sequencing data

Recessive diseases arise when both copies of a gene are impacted by a damaging genetic variant. When a patient carries two potentially causal variants in a gene, accurate diagnosis requires determining that these variants occur on different copies of the chromosome (that is, are in trans) rather than on the same copy (that is, in cis). However, current approaches for determining phase, beyond parental testing, are limited in clinical settings. Here we developed a strategy for inferring phase for rare variant pairs within genes, leveraging genotypes observed in the Genome Aggregation Database (v2, n = 125,748 exomes). Our approach estimates phase with 96% accuracy, both in trio data and in patients with Mendelian conditions and presumed causal compound heterozygous variants. We provide a public resource of phasing estimates for coding variants and counts per gene of rare variants in trans that can aid interpretation of rare co-occurring variants in the context of recessive disease.

Untangling the Context-Specificity of Essential Genes by Means of Machine Learning: A Constructive Experience

Gene essentiality is a genetic concept crucial for a comprehensive understanding of life and evolution. In the last decade, many essential genes (EGs) have been determined using different experimental and computational approaches, and this information has been used to reduce the genomes of model organisms. A growing amount of evidence highlights that essentiality is a property that depends on the context. Because of their importance in vital biological processes, recognising context-specific EGs (csEGs) could help for identifying new potential pharmacological targets and to improve precision therapeutics. Since most of the computational procedures proposed to identify and predict EGs neglect their context-specificity, we focused on this aspect, providing a theoretical and experimental overview of the literature, data and computational methods dedicated to recognising csEGs. To this end, we adapted existing computational methods to exploit a specific context (the kidney tissue) and experimented with four different prediction methods using the labels provided by four different identification approaches. The considerations derived from the analysis of the obtained results, confirmed and validated also by further experiments for a different tissue context, provide the reader with guidance on exploiting existing tools for achieving csEGs identification and prediction.

Gene essentiality in cancer is better predicted by mRNA abundance than by gene regulatory network-inferred activity

Gene regulatory networks (GRNs) are often deregulated in tumor cells, resulting in altered transcriptional programs that facilitate tumor growth. These altered networks may make tumor cells vulnerable to the inhibition of specific regulatory proteins. Consequently, the reconstruction of GRNs in tumors is often proposed as a means to identify therapeutic targets. While there are examples of individual targets identified using GRNs, the extent to which GRNs can be used to predict sensitivity to targeted intervention in general remains unknown. Here we use the results of genome-wide CRISPR screens to systematically assess the ability of GRNs to predict sensitivity to gene inhibition in cancer cell lines. Using GRNs derived from multiple sources, including GRNs reconstructed from tumor transcriptomes and from curated databases, we infer regulatory gene activity in cancer cell lines from ten cancer types. We then ask, in each cancer type, if the inferred regulatory activity of each gene is predictive of sensitivity to CRISPR perturbation of that gene. We observe slight variation in the correlation between gene regulatory activity and gene sensitivity depending on the source of the GRN and the activity estimation method used. However, we find that there is consistently a stronger relationship between mRNA abundance and gene sensitivity than there is between regulatory gene activity and gene sensitivity. This is true both when gene sensitivity is treated as a binary and a quantitative property. Overall, our results suggest that gene sensitivity is better predicted by measured expression than by GRN-inferred activity.

Inferring compound heterozygosity from large-scale exome sequencing data

Recessive diseases arise when both the maternal and the paternal copies of a gene are impacted by a damaging genetic variant in the affected individual. When a patient carries two different potentially causal variants in a gene for a given disorder, accurate diagnosis requires determining that these two variants occur on different copies of the chromosome (i.e., are in trans) rather than on the same copy (i.e. in cis). However, current approaches for determining phase, beyond parental testing, are limited in clinical settings. We developed a strategy for inferring phase for rare variant pairs within genes, leveraging genotypes observed in exome sequencing data from the Genome Aggregation Database (gnomAD v2, n=125,748). When applied to trio data where phase can be determined by transmission, our approach estimates phase with 95.7% accuracy and remains accurate even for very rare variants (allele frequency < 1×10-4). We also correctly phase 95.9% of variant pairs in a set of 293 patients with Mendelian conditions carrying presumed causal compound heterozygous variants. We provide a public resource of phasing estimates from gnomAD, including phasing estimates for coding variants across the genome and counts per gene of rare variants in trans, that can aid interpretation of rare co-occurring variants in the context of recessive disease.

Highlights from the 1st European cancer dependency map symposium and workshop

The systematic identification of tumour vulnerabilities through perturbational experiments on cancer models, including genome editing and drug screens, is playing a crucial role in combating cancer. This collective effort is known as the Cancer Dependency Map (DepMap). The 1st European Cancer Dependency Map Symposium (EuroDepMap), held in Milan last May, featured talks, a roundtable discussion, and a poster session, showcasing the latest discoveries and future challenges related to the DepMap. The symposium aimed to facilitate interactions among participants across Europe, encourage idea exchange with leading experts, and present their work and future projects. Importantly, it sparked discussions on future endeavours, such as screening more complex cancer models and accounting for tumour evolution.

Exome-wide evidence of compound heterozygous effects across common phenotypes in the UK Biobank

Exome-sequencing association studies have successfully linked rare protein-coding variation to risk of thousands of diseases. However, the relationship between rare deleterious compound heterozygous (CH) variation and their phenotypic impact has not been fully investigated. Here, we leverage advances in statistical phasing to accurately phase rare variants (MAF ~ 0.001%) in exome sequencing data from 175,587 UK Biobank (UKBB) participants, which we then systematically annotate to identify putatively deleterious CH coding variation. We show that 6.5% of individuals carry such damaging variants in the CH state, with 90% of variants occurring at MAF < 0.34%. Using a logistic mixed model framework, systematically accounting for relatedness, polygenic risk, nearby common variants, and rare variant burden, we investigate recessive effects in common complex diseases. We find six exome-wide significant (P < 1.68 × 10 - 7 ) and 17 nominally significant (P < 5.25 × 10 - 5 ) gene-trait associations. Among these, only four would have been identified without accounting for CH variation in the gene. We further incorporate age-at-diagnosis information from primary care electronic health records, to show that genetic phase influences lifetime risk of disease across 20 gene-trait combinations (FDR < 5%). Using a permutation approach, we find evidence for genetic phase contributing to disease susceptibility for a collection of gene-trait pairs, including FLG-asthma (P = 0.00205 ) and USH2A-visual impairment (P = 0.0084 ). Taken together, we demonstrate the utility of phasing large-scale genetic sequencing cohorts for robust identification of the phenome-wide consequences of compound heterozygosity.

Accurate rare variant phasing of whole-genome and whole-exome sequencing data in the UK Biobank

Phasing involves distinguishing the two parentally inherited copies of each chromosome into haplotypes. Here, we introduce SHAPEIT5, a new phasing method that quickly and accurately processes large sequencing datasets and applied it to UK Biobank (UKB) whole-genome and whole-exome sequencing data. We demonstrate that SHAPEIT5 phases rare variants with low switch error rates of below 5% for variants present in just 1 sample out of 100,000. Furthermore, we outline a method for phasing singletons, which, although less precise, constitutes an important step towards future developments. We then demonstrate that the use of UKB as a reference panel improves the accuracy of genotype imputation, which is even more pronounced when phased with SHAPEIT5 compared with other methods. Finally, we screen the UKB data for loss-of-function compound heterozygous events and identify 549 genes where both gene copies are knocked out. These genes complement current knowledge of gene essentiality in the human genome.

Recovering false negatives in CRISPR fitness screens with JLOE

It is widely accepted that pooled library CRISPR knockout screens offer greater sensitivity and specificity than prior technologies in detecting genes whose disruption leads to fitness defects, a critical step in identifying candidate cancer targets. However, the assumption that CRISPR screens are saturating has been largely untested. Through integrated analysis of screen data in cancer cell lines generated by the Cancer Dependency Map, we show that a typical CRISPR screen has a ∼20% false negative rate, in addition to library-specific false negatives. Replicability falls sharply as gene expression decreases, while cancer subtype-specific genes within a tissue show distinct profiles compared to false negatives. Cumulative analyses across tissues improves our understanding of core essential genes and suggest only a small number of lineage-specific essential genes, enriched for transcription factors that define pathways of tissue differentiation. To recover false negatives, we introduce a method, Joint Log Odds of Essentiality (JLOE), which builds on our prior work with BAGEL to selectively rescue the false negatives without an increased false discovery rate.

Cancer fitness genes: emerging therapeutic targets for metastasis

Development of cancer therapeutics has traditionally focused on targeting driver oncogenes. Such an approach is limited by toxicity to normal tissues and treatment resistance. A class of 'cancer fitness genes' with crucial roles in metastasis have been identified. Elevated or altered activities of these genes do not directly cause cancer; instead, they relieve the stresses that tumor cells encounter and help them adapt to a changing microenvironment, thus facilitating tumor progression and metastasis. Importantly, as normal cells do not experience high levels of stress under physiological conditions, targeting cancer fitness genes is less likely to cause toxicity to noncancerous tissues. Here, we summarize the key features and function of cancer fitness genes and discuss their therapeutic potential.