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

In plain English

The team generated short binding peptides, fused them to a degradation system, and screened living cells for guides that lowered target proteins.

What they found

• The authors report degraders for beta-catenin, GFAP, EWS::FLI1, and GATA2 in cell models.
• Selected guides reduced target-linked signaling or viability in the reported cell assays.

The catch

• The work is a preprint and has not been peer reviewed.
• The reported evidence comes from reporter and cell-line experiments, not animals or patients.

Methods at a glance

Study type

Pooled cell-based screening study

Biological system

Peptide-guided targeted protein degradation

Organism

Human cell lines

Dataset

Four target-specific pooled screens

Sample size

Not reported

Model architecture

Protein language models for peptide-guide generation

Training objective

Generate target-binding peptide guides

Code

Not stated in the abstract.

Data

Not stated in the abstract.

Tasks and metrics

Tasks: Reduce beta-catenin, GFAP, EWS::FLI1, or GATA2 in cells

Metrics: Target abundance; pathway activity; cell viability; apoptosis

Baselines: Target-specific reporter and endogenous-protein assays

Principal findings

• The authors report active degraders against four targets in cell models.

What remains unresolved

External validation: Not reported

• The manuscript is not peer reviewed.
• No animal or clinical validation is reported in the abstract.