Funneling modulatory peptide design with generative models: Discovery and characterization of disruptors of calcineurin protein-protein interactions

Abstract

Design of peptide binders is an attractive strategy for targeting “undruggable” protein-protein interfaces. Current design protocols rely on the extraction of an initial sequence from one known protein interactor of the target protein, followed by in-silico or in-vitro mutagenesis-based optimization of its binding affinity. Wet lab protocols can explore only a minor portion of the vast sequence space and cannot efficiently screen for other desirable properties such as high specificity and low toxicity, while in-silico design requires intensive computational resources and often relies on simplified binding models. Yet, for a multivalent protein target, dozens to hundreds of natural protein partners already exist in the cellular environment. Here, we describe a peptide design protocol that harnesses this diversity via a machine learning generative model. After identifying putative natural binding fragments by literature and homology search, a compositional Restricted Boltzmann Machine is trained and sampled to yield hundreds of diverse candidate peptides. The latter are further filtered via flexible molecular docking and an in-vitro microchip-based binding assay. We validate and test our protocol on calcineurin, a calcium-dependent protein phosphatase involved in various cellular pathways in health and disease. In a single screening round, we identified multiple 16-length peptides with up to six mutations from their closest natural sequence that successfully interfere with the binding of calcineurin to its substrates. In summary, integrating protein interaction and sequence databases, generative modeling, molecular docking and interaction assays enables the discovery of novel protein-protein interaction modulators.

In vivo selection of synthetic nucleocapsids for tissue targeting

“…1 ). It has previously been demonstrated that the incorporation of free cysteines in peptides increases tumor accumulation by forming disulfide bonds to serum albumin rather than a specific receptor-ligand interaction; therefore, we only selected miniprotein sequences without cysteines ( SI Appendix , Table S3 ) ( 27 ). Although some of the eight selected miniproteins have homologous sequences, AlphaFold2 predictions revealed unique 3D conformations, surface hydrophobic networks, and surface charge distributions ( Figs.…”

“…Notably, incorporating more rational design into both library design and in vivo selection strategy could further decrease the number of rounds of selection needed to achieve a desired result and further improve the quality of the molecules identified. For example, designing more complex surface libraries or miniprotein binder libraries against specific targets is now possible using cutting-edge machine learning-based design methods and would enable the in vivo library selection strategy to be tailored to those targets ( 27 , 38 ).…”

Pesticide chemical leads inhibiting protein-protein interactions

“…The excited fluorescent probe (i.e., SAT peptide) bound to a large molecular weight protein (i.e., OASS) emits light with a degree of polarization that is inversely proportional to the rate of molecular rotation. This fluorescence property is exploited to measure the interaction of a small labeled ligand with a larger protein and provides a basis for direct and competition binding assays 29,30 . Herein, we relied on the known SAT-derived peptide YLTEWSDYVI (SATp) conjugated with fluorescein isothiocyanate (FITC).…”