Parallel engineering and activity profiling of base editor system

Abstract

Laborious one-by-one testing of base editors has been used to select or engineer the optimal variant for inducing a specific base substitution with maximal efficiency yet minimal undesired effects. This thesis work presents a high throughput activity profiling platform to streamline the evaluation process by enabling simultaneous performance assessment of a diverse pool of base editor variant in scale. This platform generates single-nucleotide resolution readouts, allowing quantitative measurements of each variant’s performance within a cytosine base editor library, including editing efficiency, substrate motif preference, positional bias and haplotype analysis. Undesired outcomes such as impure edits, indels and noncanonical base conversions are also uncovered during the process. This work further demonstrates the discovery power of this platform via a sgRNA scaffold library, identifying two scaffold variants, SV48 and SV240, that enhance base editing efficiency while maintaining an acceptable rate of inducing undesired edits. This work also explores the potential of integrating machine learning techniques to broaden the scope of engineering with the platform, which further lowers the experimental burden. By introducing slight modifications, this platform can be adapted for parallel engineering and screening of other precise genome editors such as adenine base editors and prime editors. With the continuously expanding repertoire of genome editing tools, this platform addresses the pressing need for scalable, unbiased, and rapid benchmarking of engineered variants. This would also accelerate the development of next-generation precise genome editors and pave the way for specialised editor design by optimising, profiling, and selecting the most suitable tools for specific applications. (250 words)