Prime editor-based high-throughput screening reveals functional synonymous mutations in human cells

Synonymous mutations are generally considered neutral, while their roles in the human genome remain largely unexplored. Here we use the PEmax system to create a library of 297,900 engineered prime-editing guide RNAs and perform extensive screening to identify synonymous mutations affecting cell fitness. Unlike recent findings in yeast, group-level analyses show that synonymous mutations diverge from nonsynonymous mutations in fitness effects yet exhibit similar phenotypic distributions relative to negative controls. Following rigorous quality control, only a small subset demonstrated measurable effects. For these functional mutations, we develop a specialized machine learning tool and uncover their impact on various biological processes such as messenger RNA splicing and transcription, supported by multifaceted experimental evidence. We find that synonymous mutations can alter RNA folding and affect translation, as demonstrated by PLK1_S2. By integrating screening data with our model, we predict clinically deleterious synonymous mutations. This research deepens our understanding of synonymous mutations, providing insights for clinical disease studies.