Background: It is vital to understand healthy gut microbiota composition throughout early life stages when environments are changing, and immunity is developing. There are limited large-scale longitudinal studies classifying healthy succession of swine microbiota. The objectives of this study were to (a) determine the microbiota composition of fecal samples collected from piglets within a few days after birth until one-week post-weaning, and (b) investigate the associations of early fecal microbiota with pig growth performance in nursery and later growing stages. Fecal samples were collected from nine cohorts of 40 pigs (n = 360) from distinct farrowing sources in Ontario and Quebec, Canada at four timepoints from birth to one-week post-weaning, with pig body weight was recorded at each fecal sampling. Results: Microbiota was dominated by the phyla Firmicutes, Bacteroides and Proteobacteria. There were notable differences in genera abundance between pigs from different provinces and farming systems. Over the early life stage, the genera Bacteroides, Escherichia/Shigella, and Clostridium cluster XIVa were abundant preweaning, while Prevotella dominated post-weaning. Hierarchical clustering identified three major stages of microbiota development, each associated with distinct composition. Stage one occurs from birth to 7 days, stage two from 7 days after birth until weaning, and stage three from weaning to one-week post-weaning. Three enterotypes were identified in stage two that showed differences in growth before weaning, and in the grower production stage. Piglets with a microbiota enterotype characterized by higher abundance of Prevotella and unclassified Ruminococcaceae had lower growth performance in the pre-weaning stage, and the growing stage. Conclusion: These findings help identify the timing of microbiota shifts across early swine life which may be the optimal time for external intervention to shift the microbiota to a beneficial state. The project findings should help decrease antimicrobial use, increase animal welfare, and have positive economic impacts.
We develop a novel covariate ranking and selection algorithm for regularized ordinary logistic regression (OLR) models in the presence of severe class-imbalance in high dimensional datasets with correlated signal and noise covariates. Class-imbalance is resolved using response-based subsampling which we also employ to achieve stability in variable selection by creating an ensemble of regularized OLR models fitted to subsampled (and balanced) datasets. The regularization methods considered in our study include Lasso, adaptive Lasso (adaLasso) and ridge regression. Our methodology is versatile in the sense that it works effectively for regularization techniques involving both hard- (e.g. Lasso) and soft-shrinkage (e.g. ridge) of the regression coefficients. We assess selection performance by conducting a detailed simulation experiment involving varying moderate-to-severe class-imbalance ratios and highly correlated continuous and discrete signal and noise covariates. Simulation results show that our algorithm is robust against severe class-imbalance under the presence of highly correlated covariates, and consistently achieves stable and accurate variable selection with very low false discovery rate. We illustrate our methodology using a case study involving a severely imbalanced high-dimensional wildland fire occurrence dataset comprising 13 million instances. The case study and simulation results demonstrate that our framework provides a robust approach to variable selection in severely imbalanced big binary data.
Evaluating potential environmental and clinical impacts of industrial antibiotic use is critical in mitigating the spread of antimicrobial resistance. Using soil columns to simulate field application of swine or cattle manure and subsequent rain events, and a targeted qPCR-based approach, we tracked resistance genes from source manures and identified important differences in antimicrobial resistance gene transport and enrichment over time in the soil and water of artificially drained cropland. The source manures had distinct microbial community and resistance gene profiles, and these differences were also reflected in the soil columns after manure application. Antibiotic resistance genes (ARGs) were only significantly enriched in effluent samples following the first rain event (day 11) for both soil types compared to the control columns, illustrating the high background level of resistance present in the control soils chosen. For swine, the genes tetQ, tet(36), tet44, tetM, sul2 and ant(6)-ib persisted in the soil columns, whereas tetO, strB and sul1 persisted in effluent samples. Conversely, for cattle manure sul2 and strB persisted in both soil and effluent. The distinct temporal dynamics of ARG distribution between soil and effluent water for each manure type can be used to inform potential mitigation strategies in the future.
1 January 2025 in Canadian Journal of Microbiology
Declining costs of sequencing technology have catalyzed the widespread use of high-dimensional complex omics datasets in microbiology. While rich in information, these datasets present major analytical challenges, including sparsity, heterogeneity, and the need for robust statistical validation. Concerns about the reproducibility of findings across microbiological studies underscore the importance of standardized, transparent analytical approaches. Despite the availability of diverse statistical frameworks and machine learning methods, designing an appropriate statistical workflow (from method selection to model evaluation) remains challenging, particularly for researchers with limited advanced statistical training. Missteps in this process can lead to misinterpretation, irreproducibility, and flawed conclusions. This paper provides a structured, step-by-step framework to guide and validate the methodology of choosing the right statistical methods for both explanatory and predictive modeling in microbiology and translational research. We outline essential decision points spanning data preprocessing, feature selection, model assumptions, and model evaluation, and highlight common trade-offs and practical considerations. To demonstrate the guide’s utility, we analyze a real-world COVID-19 dataset to identify cytokine biomarkers associated with disease severity. By aligning analytical strategies with microbiology inquiry, this guide aims to enhance reproducibility, empower data-informed decisions, and promote more rigorous, interpretable research in microbiology and public health.