Data Normalization
For all multivariate analysis, positive antigens were removed (Tetanus and H1Cal2009). If any antibody feature included negative values, right-shifting was performed (by adding the minimum value for each respective feature back to all samples). Data was then log transformed to ensure that the majority of features were normally distributed, by using the following equation log10 (x+1). Values were subsequently normalized by mean centering and variance scaling by calculating respective z -scores.