Fecal Raman Spectroscopy Complements Microbiome and Metabolomic Analyses to Track Growth-Stage Biochemical Changes in Sika Deer

Background: The fecal microbiota is closely coupled to host metabolism and nutritional physiology in ruminants, but longitudinal biochemical variation across defined sika deer developmental sampling stages remains incompletely characterized. We therefore integrated microbiome, metabolomic, short-chain fatty acid (SCFA), and Raman spectroscopic measurements to describe stage-associated fecal changes. Methods: Fecal specimens were collected from male sika deer at two predefined antler-cycle stages, the rapid antler-growth stage (G; 1 March 2024) and the mature/ossification stage (M; 28 June 2024), 119 days apart. Eight animals were followed longitudinally and contributed matched specimens at both stages; two additional single-time-point specimens per stage were included, so that 16S rRNA gene sequencing, untargeted metabolomics, and SCFA quantification comprised 10 specimens per stage (n = 20), whereas Raman spectroscopy comprised the eight matched pairs (n = 16). The animal was treated as the biological unit: 16S, metabolomic, and SCFA comparisons used independent-sample tests on the 10 specimens per stage, and Raman comparisons used paired procedures on the eight matched animals, each with multiplicity correction. For Raman classification, technical spectra were averaged within each specimen and validation was structured by deer so that both stage specimens from the held-out animal were excluded from model training. Results: The two sampling stages differed in fecal microbial composition and biochemical profiles. Alpha-diversity indices did not differ significantly (all FDR-adjusted q ≥ 0.220), but genus-level analysis with FDR correction identified three differentially abundant genera (q < 0.05): UCG-005 and UCG-010 were more abundant at the G stage, and Treponema was more abundant at the M stage. Fecal butyrate was higher at the M stage (p = 0.0065, Bonferroni-adjusted p = 0.019), whereas valerate and caproate were higher at the G stage (adjusted p = 0.029 and 0.0002, respectively). Untargeted metabolomics identified 5621 differential features across the two electrospray ionization modes (|log2 fold change| ≥ 2 and p < 0.01; positive-ion features additionally selected at PLS-DA VIP ≥ 1.0), of which 2311 were increased and 3310 were decreased at the M stage relative to G, with no evidence of PLS-DA overfitting (permutation R2Y-intercept = 0.318, Q2-intercept = −0.237). Raman spectra showed stage-associated variation, with seven candidate bands showing FDR-adjusted differences (q = 0.007–0.028). A leakage-controlled classifier evaluated by grouped leave-one-deer-out cross-validation achieved an AUC of 0.875 (deer-level 95% CI 0.69–0.98; paired-label permutation p = 0.023). Conclusions: This exploratory longitudinal study identifies stage-associated variation in fecal microbial, metabolomic, SCFA, and Raman profiles in sika deer. The findings are associative rather than mechanistic because the cohort was small and developmental stage was not separated from season. Raman spectroscopy may provide complementary non-invasive biochemical phenotyping, but the classification result requires confirmation in a larger independent cohort before practical use.