Conformation-aware structure prediction of antigen-recognizing immune proteins

Area of Science:

• Structural biology
• Computational biology
• Immunology

Background:

• Accurate prediction of protein structures is crucial for understanding biological function and developing therapeutics.
• Existing models often struggle to differentiate between bound and unbound protein conformations, limiting their applicability.
• Antibodies, nanobodies, and T-cell receptors (TCRs) are key components of the immune system with significant therapeutic potential.

Purpose of the Study:

• To develop a novel structure prediction model for pan-immunoglobulins, including antibodies, nanobodies, and TCRs.
• To enable accurate prediction of both bound (holo) and unbound (apo) protein conformations.
• To improve the accuracy and efficiency of protein structure prediction for large molecules.

Main Methods:

• Introduction of Ibex, a pan-immunoglobulin structure prediction model.
• Training Ibex on labeled apo and holo structural pairs to distinguish between bound and unbound states.
• Evaluation of Ibex's performance on a benchmark of high-resolution antibody structures, assessing out-of-distribution generalization.

Main Results:

• Ibex achieves state-of-the-art accuracy in structure prediction for antibodies, nanobodies, and TCRs.
• Demonstrates superior out-of-distribution performance with a mean CDR H3 RMSD of 2.28 Å on antibody structures.
• Significantly reduces computational requirements compared to previous methods.

Conclusions:

• Ibex provides accurate prediction of both bound and unbound immunoglobulin conformations.
• The model offers a robust foundation for accelerating the design of large molecules and therapeutic development.
• Ibex represents a significant advancement in computational protein structure prediction for immunological applications.