A quality control step was applied to the metabolite data prior to analysis and described by Dervishi et al. (2021b) (link). Metabolites that were frequently (>20%) below the limit of detection or with at least 20% missing values were removed from consideration (acetic acid, isopropanol, ketoleucine, 3-hydroxyisovaleric acid). A total of 44 metabolites remained in the dataset. Data normalization (log10) of metabolite concentrations that were not normally distributed was done prior to statistical analysis. Three combinations of metabolites were also used for statistical analysis as described in Dervishi et al. (2021b) (link): 1) branched chain amino acids (BCAA), which was calculated as the sum of L-leucine, L-isoleucine, and L-valine, 2) ketogenic amino acids (ketoAA), calculated as the sum of L-lysine and L-leucine, and 3) ketones, calculated as the sum of the three ketone bodies, 3-hydroxybutyric acid acetoacetate, and acetone.
Significance of group classification, batch, enrichment, and of the covariable age at entry was determined using the following linear mixed effects model implemented in R studio (2015) statistical software:
where Υ is the trait (metabolite concentration); Group is the group classification of pigs (resilient, middle, susceptible and dead); Batch is the fixed batch effect (N = 1 to 15); NurEnrich is the fixed effect of environmental enrichment (barren or enriched); Age is the covariate of age when the pig entered the quarantine nursery; Pen is the random effect of pen by batch in the quarantine nursery; Litter is the random effect of the litter environmental effect, and e is the error. The residuals of the model were plotted and visually inspected for the presence of outliers. To test differences between group classification and enrichment conditions, the Least Square Means (LSMs) for each pair-wise comparison were estimated and their P values were adjusted using false discovery ratio correction (FDR). Differences were considered significant if P ≤ 0.05 and as a tendency towards significance if P > 0.05 and ≤ 0.10.
Significance of group classification, batch, enrichment, and of the covariable age at entry was determined using the following linear mixed effects model implemented in R studio (2015) statistical software:
where Υ is the trait (metabolite concentration); Group is the group classification of pigs (resilient, middle, susceptible and dead); Batch is the fixed batch effect (N = 1 to 15); NurEnrich is the fixed effect of environmental enrichment (barren or enriched); Age is the covariate of age when the pig entered the quarantine nursery; Pen is the random effect of pen by batch in the quarantine nursery; Litter is the random effect of the litter environmental effect, and e is the error. The residuals of the model were plotted and visually inspected for the presence of outliers. To test differences between group classification and enrichment conditions, the Least Square Means (LSMs) for each pair-wise comparison were estimated and their P values were adjusted using false discovery ratio correction (FDR). Differences were considered significant if P ≤ 0.05 and as a tendency towards significance if P > 0.05 and ≤ 0.10.
Free full text: Click here
