In Silico Directed Evolution: Engineering Humanized Peptide Transporters

We (Alper and Taner Karagöl) are engineering a new class of protein surrogates by integrating human pharmacological requirements into evolutionary stable scaffolds. This methodology provides a path to characterize complex proteins like the human intestinal peptide transporter 1 (PepT1), which is essential for oral drug delivery but difficult to study. Our approach establishes a versatile framework for creating high-fidelity models that can be applied to a wide range of pharmacological targets.

Using the 650 million parameter ESM-2 deep learning model, we performed mutational profiling to identify the energetic conflicts that typically cause engineered proteins to fail. To resolve these conflicts, our team deployed an automated, attention-guided epistatic rescue algorithm that identifies secondary mutations to stabilize the protein. This process resulted in stability improvements of up to 11,614-fold for highly problematic substitutions.

These findings were further confirmed through all-atom molecular dynamics simulations and docking calculations. This deep learning framework provides a stable platform for advanced pharmacological research and significantly reduces the time required for experimental screening.

Karagöl, A., & Karagöl, T. (2026). In Silico Directed Evolution of Humanized Peptide Transporters via Attention-Guided Epistatic Rescue.

https://doi.org/10.21203/rs.3.rs-9198655/v1