A Mammalian High-Throughput Screen for AI-Designed Peptide-Guided Protein Degraders

Targeted protein degradation (TPD) can eliminate disease-driving proteins inaccessible to conventional inhibitors, but degrader discovery remains low-throughput and poorly suited for functional selection in mammalian cells. Here, we present a high-throughput human cell-based platform for screening peptide-guided ubiquibodies (uAbs). These doxycycline-inducible degraders fuse protein language model-generated peptide guides to the CHIP$\Delta$TPR E3 ligase domain, enabling programmable TPD. Pooled uAb libraries are screened in fluorescent reporter cells by FACS and sequencing to identify enriched guides. Enriched uAbs reduced endogenous β-catenin and Wnt signaling in DLD1 cells, GFAP abundance and viability in U251 glioblastoma cells, and EWS::FLI1 abundance, EWSAT1 expression, and cell survival in Ewing sarcoma models. Screening endogenously tagged GATA2 further identified uAbs active under native genomic regulation. This platform links generative peptide design with functional mammalian selection and provides a scalable strategy for CRISPR-like proteome perturbation.