Linear discriminant analyses for the factors site, infection and colony.

R Code and the full set of LD coefficients can be found in Table S7 in File S1. Here we present only the genes with a coefficient greater than 1.1 for the site and infection effects and 2.0 for the colony effect. The magnitude of the linear discriminant coefficients indicates to what extent each factor (in this case: each gene) contributes to the predictive value of the linear discriminant function. The proportion of trace reports the predictive value of a linear discriminant function relative to the other LD functions when more than two groups are predicted and (N−1) LD functions are generated by the LDA (N being the number of groups). Leave-one-out cross validation accurately assigned samples to the correct site, infection condition, and colony 64.4%, 63.2%, and 55.2% respectively (as compared to the probabilities of 50%, 50%, and 12.5% as predicted by chance).