In general the correlation values between Ig classes were positive but few patients did show negative correlation particularly when IgE was

involved. Overall, the specificity of response to IgE poorly correlated with the other classes being less related to IgM than the others. Within the isotypes the largest amplitude in variation of correlations was observed between IgE and IgA values ( Fig. 1). In order to test whether the overall correlations between classes could be used as discriminator for classification, those coefficients for atopic and non-atopic groups of patients were compared and overall showed some differences (ANOVA, p values ranging between = < 0.001 I-BET-762 mw and 0.13). Inspection of box plots (not shown) as well as R2 and adjusted R2 values (ranging from 0.023 to 0.24 and 0.018 to 0.18, respectively) showed that although the correlations between the Ig-classes were different, they could not be

used in univariate statistic models to predict atopy. As expected when using all correlations in a multivariate approach, PLS-DA produced a reasonable predictive value for this classification (79% sensitivity and 84% specificity for prediction of atopy for left out cross-validation samples; 3 latent variables were used). Moreover, the model vectors relevant to prediction (i.e. regression and Variable Importance of Projection (VIP) vectors) produced valuable qualitative information, suggesting the expected involvement of IgE by indicating that only IgE/IgA or IgE/IgG correlation coefficients possessed some power of discrimination. In order to assess this website the feasibility of using the immunoglobulin isotypes readout directly, instead of correlation coefficients, to predict milk allergy tolerance, all readouts were used to train a PLS-DA model to discriminate between tolerant and non-tolerant subjects. The resulting model (1 latent variable, data normalized and mean centered) was able to predict tolerance with a cross-validation sensitivity and specificity of 57% and 77% respectively. Inspection of the regression vector values (result not shown) indicates that achieving tolerance is paired with a decrease in dairy

sensitivity. The spots that showed the largest variation (decreases SPTLC1 and increases) were mainly IgE and the medium contributors mainly IgA driven. Taken together and bearing in mind the clinical criteria of inclusion of the patients in this study, these results were expected; they corroborate a large number of other studies and point to specific IgE as the main parameter to be followed. In agreement with the clinical selection criteria used and as shown in Fig. 2 most of the children involved in this study have shown high levels of specific IgE to milk. Further, when clinically diagnosed milk allergic children were divided into “susceptible” and the ones that have achieved milk “tolerance” after few years, a statistically significant difference (ANOVA p = < 0.