Figure 3

Principal component analysis (PCA) of autoantibody repertoire. Multiparametric analysis by PCA allowed the comparison of repertoires of antibody reactivities of individuals in patients as compared with controls. The PCA summarizes multidimensional information (the number of dimensions is equal to the number of peaks of reactivity defined by the sections and cannot therefore be represented graphically if the number of reactivities is greater than 3) and finds all the “factors,” ie, those associations of variables (in this case the autoantibody reactivities to different protein bands) that can discriminate between groups of individuals. Factors are ranked according to their power to explain the variance between groups. PCA results are here represented in 2-dimensional graph displaying factor 1 vs factor 2, representing together 40 to 83, 6% of the total variance. At the top of each graph is displayed the number of total factors found by PCA, the percentage variance explained by factor 1 + factor 2, and the p value. Each symbol represents the repertoire of antibody reactivities of a single individual (closed circles = patients; open circles = controls).