Fig. 3. Removal of serum-specific background noise with normalization.

A Raw OD₄₅₀ readings (y-axis) from a previously published ELISA of the binding between 16 purified recombinant OspC antigens with 15 OspC variant-specific mouse sera (x-axis) [48]. Homologous bindings (solid dots) were between an antigen variant and a serum from a C3H mouse immunized with the same recombinant variant. Heterologous bindings (open dots) were between an antigen variant and a serum from a mouse immunized with a different variant. ELISA readings varied significantly among the sera (F = 51.46, p < 2.2e−16, by an ANOVA). B OD₄₅₀ readings with respect to the 15 recombinant OspC variants (x-axis), ordered by the medians. Without correcting for the serum-by-serum variability, the OspC variants did not vary significantly in reactivity with the variant-specific sera (F = 1.04, p = 0.42). C Normalized reactivity (z-score, y-axis) with respect to the variant-specific sera (x-axis). Serum-specific variability was removed (F = 0, p = 1). D Normalized reactivity (z-score, y-axis) with respect to the OspC variants (x-axis). After normalization, the OspC variants showed significant variability in reactivity with the variant-specific sera (F = 6.17, p = 8.3e−11). Five OspC variants with a median z > 0, indicating above-average reactivity, were highlighted with shaded boxes. The same five OspC variants ranked as the most reactivity without normalization. Thus, normalization did not change the ranking but greatly improved statistical confidence and precision for comparing the antigenicity among the antigen variants.