Machine learning-based development of Gadolinium binding peptides.

Area of Science:

• Biochemistry
• Molecular Biology
• Medical Imaging

Background:

• Gadolinium-based contrast agents (GBCAs) are vital for MRI but face limitations like tissue accumulation and safety concerns.
• Protein and peptide scaffolds offer selective metal binding and molecular targeting potential for improved MRI contrast agents.
• Developing short peptides with optimal gadolinium coordination and high relaxivity is a significant challenge.

Purpose of the Study:

• To design and optimize short gadolinium-binding peptide motifs using a machine learning-driven evolution platform.
• To enhance the longitudinal relaxivity (r₁) of gadolinium-based MRI contrast agents.
• To explore computational strategies for discovering novel, biologically derived contrast tags.

Main Methods:

• Utilized the Protein Optimization Engineering Tool (POET), a machine learning platform, for peptide evolution.
• Employed two algorithmic strategies: motif-based and regular-expression-based representations.
• Conducted two rounds of directed evolution and experimental screening of 74 initial EF-hand derived peptides.

Main Results:

• Identified peptides with a 24% increase in r₁ ratio and a 55% improvement in absolute r₁ compared to controls.
• Found that higher relaxivity peptides typically have a more negative net charge and lower isoelectric point.
• Observed selective enrichment of acidic and small polar residues (Asp, Gly, Thr) and depletion of bulky hydrophobic/basic residues.

Conclusions:

• Demonstrated a generalizable framework integrating computational evolution and biophysical screening for discovering new Gd-binding motifs.
• This approach enables the engineering of responsive, tunable, and biocompatible MRI contrast tags.
• The developed peptides offer a scalable route for precision imaging and molecular diagnostics.