Statistical analyses of laboratory ’validation experiments

We used generalized linear mixed models (GLMMs; Zuur et al., 2013) to evaluate the effects of antibody dilution, protein quantity, time stored at ambient temperature, tissue preservative and skin sampling location. The relationships between response (mean protein expression) and independent variables were modelled using a γ distribution and log link because protein expression values are always positive and usually distributed with a positive skew. For expediency, we modelled relationships on the basis of the four functional groups rather than the 31 individual proteins, except for the antibody dilution experiment, which focused on the expression of a single protein (cytokeratin). Although this assumed that each protein within a functional group would respond in the same manner, we recognized that this might not always be the case. Protein was therefore also included in models as a fixed factor in addition to the potential effects of interest (i.e. antibody dilution, protein quantity, etc). The source of skin samples and individual identity of bears were included in the models as random effects. We used the ‘glmer’ function in package ‘lme4’ (Bates et al., 2014) in R 3.1.2 (R Core Team, 2014) for model development. For models in which the potential effect of interest was significant (P ≤ 0.05), we compared mean protein expression among all possible pairs by Tukey’s HSD (honest significant difference) test using the ‘glht’ function in package ‘multcomp’ (Hothorn et al., 2008) in R 3.1.2 (R Core Team, 2014).