Medicinal Chemistry & Chemical Biology, Short talk
MC-015

Discovery of de novo peptidic binders for undruggable proteins

Y. Kamei1,2, M. Pacesa3, G. Grammbitter1, G. Menoud4, L. Nickel3, P. Franz1, C. Heinis4, B. Correia3, J. Waser2*, B. Fierz1*
1Laboratory of Biophysical Chemistry of Macromolecules, Institute of Chemical Sciences and Engineering, École Polytechnique Fédérale de Lausanne, 2Laboratory of Catalysis and Organic Synthesis, Institute of Chemical Sciences and Engineering, École Polytechnique Fédérale de Lausanne, 3Laboratory of Protein Design and Immunoengineering, School of Engineering, École Polytechnique Fédérale de Lausanne and Swiss Institute of Bioinformatics, 4Laboratory of Therapeutic Proteins and Peptides, Institute of Chemical Sciences and Engineering, École Polytechnique Fédérale de Lausanne

Dysregulation of transcription often leads to carcinogenesis and cancer proliferation, making chemical modulation of chromatin factors a potential therapeutic strategy. However, protein-protein interactions (PPIs) are deemed undruggable by small molecules due to the lack of well-defined binding pockets. Peptides have emerged as a suitable modality for targeting such undruggable proteins by mimicking PPIs and engaging flat surfaces that are typically inaccessible to small molecules.

While artificial intelligence (AI) has enabled the generation of high-affinity protein binders1, the design of peptidic binders by AI remains largely underexplored. To date, peptidic binders are typically identified by screening massive libraries (>108 members) via display technologies2 or high-throughput synthesis3.

Here, we describe a platform for the de novo design and rapid validation of peptidic binders that target typically undruggable proteins, including intracellular oncogenic chromatin factors. Our AI-designed peptides were synthesized by solid-phase peptide synthesis in a 384-well format and screened by affinity-selection mass spectrometry. We successfully identified binders for challenging targets. Our integrated platform may serve as a broadly applicable approach for targeting undruggable proteins.

[1] Martin Pacesa, Sergey Ovchinnikov, Bruno E. Correia et al., Nature 2025, 646, 483-492.
[2] Toby Passioura, Takayuki Katoh, Yuki Goto, Hiroaki Suga, Annu. Rev. Biochem. 2014, 83, 727-752.
[3] Bradley L. Pentelute et al., Nat. Commun. 2020, 11, 3183.