Аннотация
Objectives Links between gut microbiota and insulin resistance (IR) vary across populations. We profiled the fecal microbiota of Kazakh adults to test whether community composition associates with IR at broad (phylum → genus) and species levels. Methods In a cross-sectional case control study (N = 200; IR = 183, controls = 17), TyG indexed IR status. 16S rRNA sequencing (two primer pools; nine hypervariable regions) characterized taxa. After CSS normalization, we compared presence/absence across groups (χ2) and modeled species with univariate and multivariable logistic regressions, using absence of each species as the predictor. Results High-level composition did not differ between IR and controls (phylum, class, family, genus; all p > 0.05). In contrast, several species differed. In univariate models, absence of Actinomyces odontolyticus (OR = 25.55, p = 0.010), Bifidobacterium kashiwanohense (OR = 12.69, p = 0.015), Lactobacillus sp. (OR = 5.71, p = 0.020), and Streptococcus lactarius (OR = 6.27, p = 0.044) associated with higher IR odds, suggesting protection when present; whereas absence of Alistipes onderdonkii (OR = 0.30, p = 0.044) and Prevotella copri (OR = 0.19, p = 0.003) associated with lower IR odds, suggesting risk when present. In multivariable models, these signals persisted: absence of P. copri (OR = 0.146, p = 0.003) and Roseburia inulinivorans (OR = 0.143, p = 0.011) predicted lower IR odds (risk alignment), while absence of Lactobacillus sp. (OR = 8.29, p = 0.016) and Coprococcus catus (OR = 7.04, p = 0.004) predicted higher IR odds (protective alignment). Conclusion In this Kazakh cohort, no broad compositional signal emerged, but species-specific associations were strong and bidirectional. Findings highlight population-specificity and identify candidate species associated with IR that may serve as hypothesis-generating targets for future validation. Any attempt to modulate these taxa for insulin resistance is unproven and requires function-resolved, diet-measured longitudinal studies and randomized trials before clinical application. The IR:control imbalance (183:17) increases uncertainty for low-prevalence taxa; species-level findings are hypothesis-generating and require validation in a more balanced design. Because 16S rRNA profiling does not measure gene functions or metabolites, these species–IR associations are hypothesis-generating and warrant validation using shotgun metagenomics and metabolomics.