This thesis addresses limitations of current single-cell sequencing technologies and proposes alternative experimental designs to increase statistical power.

Single-cell sequencing enables high-throughput and high-dimensional studies of bio- logical systems. This is particularly useful in functional genomics screens that introduce perturbations to investigate and reconstruct the regulatory networks within and between cells. CRISPR screens are the leading method of conducting functional genomics screens due to their specificity, precision and ease of use. Single-cell sequencing and CRISPR screens have recently been integrated to create the first generation of high-throughput and high-dimensional functional genomics screens.

However, the development of efficient experimental design is lagging behind technical advances for single cell CRISPR (scCRISPR) screens. To address this challenge, I developed two wet-lab-aware statistical simulators to compare various experimental protocols and assess their performances. I specifically studied (1) how different protocols affect the performance of scCRISPR screens and (2) how to reduce overall sparsity in the data.

First, to increase the statistical power of scCRISPR screens, new alternative experimental protocols need to be investigated. However, conducting these experiments is time consuming and expensive; therefore, I developed crisprPower, a statistical simulator capable of simulat- ing scRNA-Seq CRISPR screens and allowing researchers to investigate alternative protocols. Simulations showed that the current experimental design of scRNA-Seq CRISPR screens is underpowered, requiring at least 600 cells to observe the effect of a perturbation compared to targeted panels that only need 100 cells.

Second, I proposed an improved single-cell experimental protocol that decreases sparsity. I developed a new simulator, Minerva, to simulate the effects that experimental enrichment protocols (e.g. antibody pulldowns or PCR) would have upon the observed counts of single- cell datasets. Using Minerva, I showed that it is possible to reduce the sparsity of single-cell datasets and measure lowly expressed genes that would have been lost otherwise. My research shows that improving the experimental design of single-cell protocols using theoretical analysis and simulation leads to concrete and easily implemented recommendations that improve scRNA-Seq CRISPR screens and single-cell sequencing methods.