Fig. 4: Model predictions and experimental evaluation of Pareto optimal affinities and specificities for emibetuzumab mutants in the sorted antibody libraries. | Nature Communications

Fig. 4: Model predictions and experimental evaluation of Pareto optimal affinities and specificities for emibetuzumab mutants in the sorted antibody libraries.

From: Co-optimization of therapeutic antibody affinity and specificity using machine learning models that generalize to novel mutational space

Fig. 4

A Visualization of the predicted Pareto frontier for the 4000 library variants using the LDA models reported in Fig. 3. Antibody mutants located along the Pareto frontier are predicted to display co-optimized affinity and specificity, as increases in either property requires decreases in the other one. B, C Antibody mutants (41 antibodies) at or near the predicted Pareto frontier were generated as soluble IgGs, and their (A) antigen binding (1 nM HGFR) and (B) non-specific binding (0.1 mg/mL ovalbumin) were evaluated and plotted relative to the continuous model predictions. D Comparison of the experimentally measured values of antigen and non-specific binding for the 41 IgGs reported in parts (B) and (C). In (B–D), the experimental measurements are averages of three independent repeats, and the error bars are standard deviations. Independent two-sided t-tests were performed to determine significance.

Back to article page