Engineering multispecific antibodies with complete killing selectivity through the closed-loop integration of machine...

Area of Science:

• Immunology
• Biotechnology
• Computational Biology

Background:

• On-target, off-tumor toxicities limit T-cell engager (TCE) efficacy in solid tumors.
• Developing potent and tumor-selective TCEs requires exploring vast design spaces.

Purpose of the Study:

• To introduce EVA™, a closed-loop design platform for optimizing T-cell engagers.
• To demonstrate EVA™'s capability in identifying potent and tumor-selective TCE candidates.

Main Methods:

• EVA™ integrates high-throughput functional assays with multi-objective Bayesian optimization.
• Iterative design-build-test-learn cycles were used to explore combinatorial TCE spaces.
• A HER2×CD3 case study analyzed 44,160 designs based on valency, topology, affinity, and spacing.

Main Results:

• EVA™ achieved a 14-fold enrichment of potent, tumor-selective candidates compared to a Sobol baseline.
• Optimized TCEs demonstrated sub-10 pM potency against HER2-high cells with near-complete efficacy.
• Selectivity over HER2-low models exceeded 10,000-fold, aligning with avidity gating principles.

Conclusions:

• EVA™ provides an efficient, data-driven approach for rapid T-cell engager optimization.
• Density-gated avidity emerges as a key design principle for enhancing TCE safety and efficacy.
• The platform successfully generalized to a second target, showcasing its broad applicability.