Computational methods to evaluate and interpret genome-wide perturbation screens
2021-08
Loading...
Persistent link to this item
Statistics
View StatisticsJournal Title
Journal ISSN
Volume Title
Title
Computational methods to evaluate and interpret genome-wide perturbation screens
Authors
Published Date
2021-08
Publisher
Type
Thesis or Dissertation
Abstract
Genome-wide perturbation screens are a powerful tool to learn about biological systems. They allow us to systematically delete or mutate individual genes or combinations of genes, measure the impact of these perturbations, and learn how biological systems are functionally organized. This has been powerfully demonstrated in the model organism yeast, where all individual genes were knocked out, and the effect of gene deletion on yeast was methodically studied. In the last decade, our lab has been involved in multiple major efforts to knock out combinations of genes in yeast and showed its efficacy at uncovering functional relationships between genes. In the past few years, many different groups (including ours) have been undertaking similar efforts in human cells using pooled CRISPR/Cas9 screening approaches. These endeavors have produced a wealth of genome-wide perturbation datasets. In this dissertation, we focus on the development of computational methods to benchmark and interpret genome-scale perturbation datasets from yeast to humans. We begin with a discussion of methods for interpreting higher-order genetic interactions in yeast. We first extend this by exploring the impact of environmental perturbation on genetic interactions using 14 different chemical conditions. This study highlights the robustness of the global reference genetic interaction network in yeast, as the functional rewiring in the presence of changing environments is rare and less functionally informative. Next, we describe methods for scoring high-throughput trigenic interaction experiments in yeast. This method and associated software tool enables quantification of higher-order interactions involving triple mutant combinations and was used to map the first large-scale network of higher-order interactions in any species. The latter half of the thesis focuses on computational approaches for generating, scoring, and interpreting the results of genome-wide perturbation screens in human cells. One important aspect of this is to be able to systematically evaluate and compare different datasets and associated methods. To this end, we develop a method named FLEX to interpret functional information in a CRISPR screen dataset and systematically compare different competing datasets or methods. We employ FLEX on a genome-wide single knockout dataset, the DepMap data, and demonstrate that it reveals a major functional bias for mitochondrial genes, which we hypothesize is related to protein stability. Another focus of our work on methods for interpreting CRISPR screens is to develop approaches that can help to quantify the reproducibility of these perturbation screens. Specifically, we developed a method called JEDER that estimates error rates for CRISPR screens and establishes a way to evaluate replicated CRISPR screens without knowledge of any external gold standard. We highlight the importance of reporting relevant reproducibility metrics by demonstrating the increased difficulty in reproducing differential effects (e.g. genetic interactions) as compared to primary effects (e.g. single mutant fitness). Given accurate methods for scoring and quality control of CRISPR screens, these technologies can be applied to map large-scale genetic interaction maps for human cells. In the final chapter, this thesis describes the results of our computational analysis of the first genome-wide interaction network for human cells. We establish a set of genomic features that relate to gene essentiality and evaluate how different functional standards and genomic or proteomic data relate to different types of interactions. Finally, we summarize different functional neighborhoods and how well they are captured by the current genetic interaction map and suggest approaches to drive future interaction screening efforts.
Description
University of Minnesota Ph.D. dissertation. August 2021. Major: Computer Science. Advisor: Chad Myers. 1 computer file (PDF); xiv, 186 pages + 2 supplemental files.
Related to
Replaces
License
Collections
Series/Report Number
Funding information
Isbn identifier
Doi identifier
Previously Published Citation
Suggested citation
Rahman, AHM Mahfuzur. (2021). Computational methods to evaluate and interpret genome-wide perturbation screens. Retrieved from the University Digital Conservancy, https://hdl.handle.net/11299/243115.
Content distributed via the University Digital Conservancy may be subject to additional license and use restrictions applied by the depositor. By using these files, users agree to the Terms of Use. Materials in the UDC may contain content that is disturbing and/or harmful. For more information, please see our statement on harmful content in digital repositories.