Abstract The aim of this study was to identify which omics data blocks among metabolomic applied on plasma or fecal samples and metagenomic of feces best explain differences between piglets reared in two production chains, and which blocks capture the greatest variance across growth stages. At 27 ± 2 days of age, 40 piglets from 17 sows were selected, balanced by body weight and litter at weaning, and assigned to two swine production chains (SPC1 and SPC2) characterized a priori by productivity, antibiotic use, and infection occurrence. Each chain comprised one weaning unit and one growing–fattening unit, including a total of 4 commercial farms. Blood and feces were collected from all pigs at T1 (two weeks post-weaning), T2 (seven weeks post-weaning), T3 (about two weeks after transfer to growing–finishing units), and T4 (slaughter). Fecal samples underwent shotgun metagenomic sequencing to quantify microbial species and functions (KEGG orthologs, KOs) with marker gene–based pipelines (MetaPhlAn 4 and HUMAnN 3) and untargeted metabolomics (∼1,200 metabolites; Metabolon Inc.), while blood provided plasma metabolomics (∼1,200 metabolites; Metabolon Inc.), hematology, and markers of oxidative stress and inflammation. At each time point, multi-block sparse PLS-DA (DIABLO) models contrasting SPC1 vs SPC2 were fitted with repeated cross-validation to tune components and block-specific features; discrimination was summarized as error rate (mean ± SD), and block-wise variance explained was derived from component-level R²/Q². Overall (weighted-vote) error rates were 10% (±12) at T1, 7% (±4) at T2, 12% (±7) at T3, and 25% (±8) at T4. The most discriminative single data block varied by stage when expressed as error rate: plasma metabolites at T1 (11% ± 11) and T2 (2% ± 2), blood formula and inflammatory markers at T3 (11% ± 4), and fecal metabolites at T4 (26% ± 7). In contrast, KOs consistently explained the largest share of variance: 0.21 (±0.22) at T1, 0.11 (±0.17) at T2, 0.12 (±0.12) at T3, and 0.11 (±0.02) at T4, while explained variance of blood formula increased at mid–late stages (T3: 0.10 ± 0.05; T4: 0.12 ± 0.08) and plasma/fecal metabolomes contributed smaller but stable portions (plasma T2: 0.09 ± 0.01; feces T3: 0.08 ± 0.05). These results indicate that the data blocks explaining the most variance are not always the best at distinguishing SPCs: both KOs functions and species capture broad variability shared by both SPCs, which weakens their discriminative power. Plasma metabolites may provide a more reliable proxy for short-term physiological adaptation to environmental stimuli, whereas microbial composition and functional potential tend to change more slowly.
Trevisi et al. (Wed,) studied this question.