Measuring error rates in genomic perturbation screens: gold standards for human functional genomics.
Hart T, Brown KR, Sircoulomb F, Rottapel R, Moffat J. · 2014
Read Measuring error rates in genomic perturbation screens: gold standards for human functional genomics. by Hart T, Brown KR, Sircoulomb F, Rottapel R, Moffat J. (2014) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc4299491.txt.
Introduction
Traver Hart, Kevin R Brown, Fabrice Sircoulomb, Robert Rottapel, and Jason Moffat built reference lists of genes that human cells can and cannot live without. Their paper appeared in Molecular Systems Biology in 2014. Genome wide screens knock down or knock out genes one by one to find which ones a cell needs, but stray effects and unmeasured error rates clouded the results.
From many screens the team derived gold standard sets of essential and nonessential genes. These sets let labs score how accurate a screen is and compare RNA interference with the newer CRISPR method on firm ground. Genes needed only by cancer cells stand out as drug targets, so cleaner screens sharpen that search as well.
Catalog record
| CREATOR | Hart T, Brown KR, Sircoulomb F, Rottapel R, Moffat J. |
|---|---|
| YEAR | 2014 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 79 KB |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Measuring error rates in genomic perturbation screens: gold standards for human functional genomics By Traver Hart, Kevin R Brown, Fabrice Sircoulomb, Robert Rottapel, Jason Moffat Published in: Molecular Systems Biology DOI: 10.15252/msb.20145216 ABSTRACT Abstract Technological advancement has opened the door to systematic genetics in mammalian cells. Genome-scale loss-of-function screens can assay fitness defects induced by partial gene knockdown, using RNA interference, or complete gene knockout, using new CRISPR techniques. These screens can reveal the basic blueprint required for cellular proliferation. Moreover, comparing healthy to cancerous tissue can uncover genes that are essential only in the tumor; these genes are targets for the development of specific anticancer therapies. Unfortunately, progress in this field has been hampered by off-target effects of perturbation reagents and poorly quantified error rates in large-scale screens. To improve the quality of information derived from these screens, and to provide a framework for understanding the capabilities and limitations of CRISPR technology, we derive gold-standard reference sets of essential and nonessential genes, and provide a Bayesian classifier of gene essentiality that outperforms current methods on both RNAi and CRISPR screens. Our results indicate that CRISPR technology is more sensitive than RNAi and that both techniques have nontrivial false discovery rates that can be mitigated by rigorous analytical methods. Keywords: cancer, CRISPR, essential genes, RNAi, shRNA INTRODUCTION In the early 1900s, Lucien Cuénot observed unusual patterns of inheritance when studying coat color in mice and, from his many crosses, never produced a single homozygous yellow mouse (Cuenot, 1905; Paigen, 2003). Not long after these observations, it was shown that Cuenot's crosses resulted in what appeared to be non-Mendelian ratios because he had discovered a lethal gene (Castle & Little, 1910). W.E. Castle and C.C. Little demonstrated that one-quarter of the offspring from Cuénot's crosses between heterozygotes died during embryonic development, ushering in embryonic lethality, or death, as a new phenotypic class for geneticists (Castle & Little, 1910). Consequently, the idea that organisms harbor sets of lethal or essential genes has taken shape over the past century. In the past dozen years or so, systematic genomic studies in eukaryotic model systems have defined sets of lethal or essential genes under defined growth conditions, providing a nexus for biologists to study the essential molecular processes that occur during cell growth and proliferation. The importance of defining essential genes is threefold. First, it provides a blueprint for all components necessary for a cell to grow and divide under defined conditions. Second, it provides a parts list that can be deconstructed to uncover all the necessary cellular and molecular functions that proceed during cell growth and division under defined experimental conditions. Third, the list of essential genes and related functions provides a reference point for understanding disease. Indeed, the accurate identification of human disease genes is among the most important goals of biomedical research, and there exists a complex relationship between disease genes and essential genes, particularly for cancer genomes. For example, a recent analysis has shown that the cumulative effects of copy number variants of cancer drivers and essential genes along a chromosome explain the recurring patterns of somatic copy number alterations of whole chromosomes and chromosome arms in cancer genomes (Davoli et al, 2013). Broadly speaking, a gene is defined as essential if its complete loss of function results in a complete loss of fitness. In single-celled organisms, this is a fairly straightforward assessment; however, in metazoans, a gene could be reasonably classified as essential if its loss of function resulted in sterility or failure to develop to adulthood. In practice, a prenatal lethal phenotype is typically the criterion for essentiality. Given the absence of a set of well-established human essential genes, researchers have generally relied on orthology to infer essentiality. Lethal or essential gene sets have been generated under defined growth conditions for a number of eukaryotic model systems including the budding yeast S. cerevisiae (Winzeler et al, 1999; Giaever et al, 2002), the fission yeast S. pombe (Kim et al, 2010), C. elegans (Kamath et al, 2003), D. melanogaster (Boutros et al, 2004; Dietzl et al, 2007), M. musculus (White et al, 2013) and others. Across model organisms, essential genes are more likely to be hubs in protein-protein interaction networks (Jeong et al, 2001), a phenomenon driven to some degree by membership in large essential protein complexes (Hart et al, 2007; Zotenko et al, 2008). Moreover, model organism essentials are less likely to have paralogs (Makino et al, 2009), consistent with the model of gene duplication buffering loss-of-function phenotypes (Gu et al, 2003). Human orthologs of mouse knockouts which give rise to developmental lethal phenotypes are themselves enriched for developmental disease genes, even above the bias toward developmental genes in the mouse knockout set (Makino et al, 2009). Furthermore, ubiquitously expressed human genes are very likely to contain a large proportion of essential genes and are different in their evolutionary conservation rates (i.e. higher nonsynonomous/synonomous substitution rates), DNA coding lengths, and gene functions compared with disease genes and other genes (Tu et al, 2006). Experimental assays of human gene essentiality are performed in cell lines. RNA interference has, to-date, been the weapon of choice for genome-scale fitness screening, with roughly two hundred published cell line screens (Luo et al, 2008; Schlabach et al, 2008; Silva et al, 2008; Cheung et al, 2011; Marcotte et al, 2012). Other approaches include gene traps in haploid human cells (Carette et al, 2009; Burckstummer et al, 2013) and, more recently, genome-scale gene editing approaches using lentiviral-based CRISPR technologies (Shalem et al, 2013; Wang et al, 2013). The RNAi screens to-date have typically been conducted in cancer cell lines or normal counterparts to elucidate not only which genes are essential, but also which genes are differentially essential in different contexts, with the ultimate goal of identifying genes or pathways that are tissue-, subtype-, or even tumor-specific (i.e. genotype- or context-dependent essential or lethal genes). With the widespread adoption of pooled library shRNA screens has come the understanding that there are caveats to this type of genetic screening approach. In particular, off-target effects can lead to false positives (Echeverri et al, 2006; Moffat et al, 2007), if the unintended target of an shRNA hairpin is an essential gene. To mitigate these effects, analytical approaches have been developed that look for phenotypic consistency across multiple hairpins targeting a gene (Luo et al, 2008; Cheung et al, 2011; Marcotte et al, 2012) and among the same hairpins in different screens (Shao et al, 2013). Not surprisingly, different approaches can yield different results, and the degree to which false positives contaminate results is largely unknown (Kaelin, 2012). No method currently exists to systematically evaluate these various approaches. Studies in other areas of functional genomics have relied on ‘gold-standard’ positive and negative reference sets (Jansen & Gerstein, 2004) to evaluate the sensitivity and specificity of, for example, protein-protein interactions (Hart et al, 2006; Havugimana et al, 2012). This approach applies equally well to gene essentiality studies, where negatives can outnumber positives by an order of magnitude. However, no such gold standards currently exist for screens using mammalian, and more specifically human, cell lines. The developmental essentials inferred by orthology contain many genes which are, by definition, essential for whole-organism development